481 Matching Annotations
  1. Oct 2020
    1. We have: defined procurement principles and standards. These are (in summary): 100% renewable energy and/or carbon neutral suppliers 0% to landfill and an annual increase in reuse and material recycling increased transparency across HMG, suppliers and the supply chain 100% traceability of ICT at end of life a yearly increase in procured ICT and services that is remanufactured/refurbished

      Pretty explicit

    1. I'm really curious whether or how creating more effective car battery recycling technologies could support or underwrite consumer electronics recycling. For years, the line on e-waste has been that the economic model for electronics recycling just doesn't work. As devices get smaller, extraction gets more time-consuming and with increased component miniaturization, recycling companies are getting less and less value from what metal they can recover (although notably, Redwood Materials is refining their battery recycling process through recycling consumer electronics). Recycling car batteries at scale fills a massive need and companies will definitely pay for those materials. If Redwood can get it right, they could potentially make a less-profitable recycling niche more viable with the resources afforded by its primary more profitable niche.

      Basically, ways to reclaim existing materials in the technosphere, than need to mine stuff

  2. Sep 2020
    1. In addition, the recent need to accelerate deep-learning and artificial intelligence applications has led to the emergence of specialized accelerator hardware, including graphics processing units (GPUs), tensor processing units (TPUs) and field-programmable gate arrays (FPGAs). Owing to its in-memory data model, NumPy is currently unable to directly utilize such storage and specialized hardware. However, both distributed data and also the parallel execution of GPUs, TPUs and FPGAs map well to the paradigm of array programming: therefore leading to a gap between available modern hardware architectures and the tools necessary to leverage their computational power.

      Ah, so it whie it supports SIMD, it doesn't support this stuff yet.

    2. NumPy operates on in-memory arrays using the central processing unit (CPU). To utilize modern, specialized storage and hardware, there has been a recent proliferation of Python array packages

      So these can drop down to take advantage of SIMD and all that?

  3. Aug 2020
    1. his dream of it being as easy to “insert facts, data, and models in political discussion as it is to insert emoji” 😉 speaks to a sort of consumerist, on-demand thirst for snippets, rather than a deep understanding of complexity. It’s app-informed, drag-and-drop data for instant government.
  4. Jun 2020
    1. James Hetherington, Directorate of Digital Research Infrastructure, UK Research and Innovation

      Holy shit. AMEE James? Small world.


    1. The set-up of a cloud services marketplace for EU users from the private and public sector will be facilitated by the Commission by Q4 2022. The marketplace will put potential users (in particular the public sector and SMEs) in the position to select cloud processing, software and platform service offerings that comply with a number of requirements in areas like data protection, security, data portability, energy efficiency and market practice.

      Best source of compute, but across the entire EU?

  5. May 2020
    1. In France, as in other Western countries, for various reasons too long to mention in this study, the collection level of waste electrical and electronic equipment is around 45%.

      This is higher than I thought - almost half of all electronics are collected in France?

    2. Had they been implemented as of 2010, these 4 measures would have reduced the global digital footprint over the period observed (2010 to 2025) by between 27% and 52%

      How would you back this claim up, or disprove it? It feels like catnip for journalists, but I don't see how you could interrogate it.

    3. Reducing the number of flat screens by replacing them with other display devices such as augmented / virtual reality glasses, LED video projectors, etc.

      Presumably this would be down to the energy intensive nature of making a flatscreen?

    4. Their contribution to the impact of the digital world thus goes f rom less than 1% (all environmental indicators combined) in 2010 to between 18% and 23% in 2025. It is huge!

      These numbers seem comically high

    5. In 2025, user equipment will concentrate from 56% to 69% of the impact. For example, 62% of digital GHG emissions will be user-related, 35% of which comes from equipment manufacturing.

      What kind energy mix does this assume in 2025?

    6. The hierarchy of impact sources is as follows, in descending order of importance:1. Manufacturing of user equipment;2. Power consumption of user equipment;

      So these guys say it's the end use that's the big problem, not the infrastructure or network usage per se.

    7. In 2019, the mass of this digital world amounts to 223 million tonnes, the equivalent of 179 million cars of 1.3 tonnes (5 times the number of cars in France).

      So this might be a bit like the idea of biomass, but for technology?

    1. We've likely cleared half the biomass of plants on earth, and after that, they make up more than 80% of all the biomass of everything alive today.

      The livestock we farm is likely around 10 times the biomass of all wild mammals and birds

    1. This document gives a good background on where you might choose to use LCA, and where it's not such a useful tool.

      It's from 2012, but what it's saying is mostly in line with my understanding of the subject.

      Things have moved on since it was written and the open source projects seem not to be all that active now.

    2. © Intellect July 2012

      This report is 8 years old.

    3. Source: IBM/Carnegie Mellon University Carbon Footprint Research Study

      What year is this from?

    4. Analysis of a web-search of published LCA study results for ICT devices showing percentage use-stage carbon. Source: Darrel Stickler, Cisco

      This diagram is v handy - at a glance, it gives an idea where the main levers for reducing might be depending on the kind of product

    5. This framework tends to follow a similar pattern which includes most or all of the following steps: set goals and define scope, inventory analysis, impact assessment, interpretation, reporting and critical review.

      This is a really helpful diagram for explaining what the alphabet soup of standards means, and how you might apply them.

    6. Even the OECD definition of ICT includes a whole range of consumer electronic (CE) products that many would not expect to see classed as ICT.

      Where is the official definition in the OECD?

    7. And with carbon, as with calories, it is the pies –or rather their carbon equivalents -that are important. LCA is a poor tool for differentiating between strawberries and raspberries, but it is a wonderful tool for identifying where the pies are, and who is eating them

      This is totally worth using in future


    1. 1,211,22

      This is the final figure for scope 1, 2, and 3, after accounting for market based figures for CO2 from energy.

      They've split out a huge chunk of emissions in scope 3 as *other, but it's not immediately obvious to me what this includes, and they've chosen not to account for it when it comes to purchasing the offsets anyway.

    2. 15,027,224

      So this figure is actually pretty close to Microsoft now. The figures I was looking at were for just scope 1 and 2, not scope 3.

    1. As discussed above, this represents approximately 5% of ExxonMobil’s anticipated Permian oil production in that year and could add as much as 3.4 million metric tons of carbon into the atmosphere by 2025

      Oh so it wasn't per year, after all. It's over 5 years, so I thikn this is 680,000 tonnes of CO2 each year instead, of 3.4m tonnes each year.

    2. Displace dirty energy demand: New renewables supply displaces demand for existing dirty electricity generation.

      How do you demonstrate this at a supplier level.

    3. The Oil Climate Index from the Carnegie Institution gives full lifecycle emissions for a selection of global crude oils. That analysis does not specifically profile Permian crude oils, but does find that the median U.S. crude oil will lead to 0.51 metric tons of CO2-eq over its lifecycle.[

      Are these figures the updated ones where life cycle figures for fossil fuels turn out to be higher than previously thought?

    4. Public commitment to no longer offer machine learning or high performance computing capabilities for the oil and gas sector for the purpose of new exploration or increased production, and to not renew existing contracts. End membership with the Open Subsurface Data Universe Forum.

      If you were a company doing this, what would the wording look like?

    5. In 2018, Google attracted former President and General Manager of BP, Darryl Willis, as VP of Oil, Gas, and Energy at Google Cloud, where he was tasked with developing new products and solutions and building trusted relationships with key leaders and companies in the oil and gas sector

      VP of Oil, Gas and Energy

    6. In addition to specific machine learning contracts for managing upstream datasets, all three cloud companies are part of the Open Group Open Subsurface Data Universe™ (OSDU) Forum, an international group of oil and gas operators, including Shell, Chevron, Schlumberger, ExxonMobil, Total, BP, ConocoPhillips and Devon Energy, that aims to build an open data platform for the oil and gas industry.

      Open Data for Evil

    7. o realize the climate commitments they have set, Google, Microsoft, and Amazon must continue to reduce carbon emissions throughout their own operations and publicly distance themselves from customers that are making the climate crisis worse.

      It's crazy that this would be seen as controversial, and yet…

    1. The action was the digital equivalent of queueing up at McDonalds and ordering the non-existent vegan, zero-waste Happy Meal again and again. Rebels targeted a different polluter each day, including fossil fuel companies Shell and BP, shipping company Maersk, and the Danish Finance Ministry for its recent bailout of Scandinavian Airlines.

      This sounds a lot like DDOSing websites to me.

    1. What exactly has changed?

      Y'kow, they could have mentioned this earlier in this long post full of verbiage.

    1. Technological improvements also play a role in driving down battery costs. Frith points out that the term lithium-ion battery is actually an umbrella term for a number of different battery chemistries.

      Did not know this.

    1. She labelled this phenomenon “socially organised denial”.

      This is a useful term

    2. With the backing of Green Alliance and some philanthropic funders, I set up a training programme. We offered parliamentary candidates and new MPs the chance to learn about the science, policy and politics of climate in a series of tailor-made workshops. We worked with small groups of around 10 politicians, all from the same party, to allow them to question and debate freely.

      I wonder how much this cost, to design these kinda of workshops?

    1. Strengthening: The team understands its role in the larger organizational system and actively works to make that system more successful.

      Ah, so strengthening refers to an outward use of the term, not the team itself getting stronger per se.

    2. Although you may be tempted to hire new employees to fill the gaps, it’s usually more effective to include employees who already understand your business’s unique priorities and constraints.

      So, basically they propose embedding a domain expert in an existing team, as it's easier to propose than hiring a new role

    3. team chartering


    4. Each fluency zone brings new benefits, so it may seem that the model should be treated as a maturity model, in which the goal is to reach maximum maturity. That would be a mistake. Unlike maturity models, where more mature is always better, the fluency model describes a collection of choices. Each zone represents a fully-mature choice. Each one brings value.

      Using fluency instead of maturity - less of a value judgement, and more of a deliberate decision. no one wants to admit to being immature, but admitting to not being fluent is easier

    5. Although teams develop proficiencies in any order, even from multiple zones simultaneously, we’ve observed that teams tend to gain zone fluency in a predictable order.

      So, there's a fairly clear order of where to start.

    6. Focusing teams produce business value. Delivering teams deliver on the market cadence. Optimizing teams lead their market. Strengthening teams make their organizations stronger

      Focussing, Delivering, Optimising, Strengthening. How

    1. Once done with injecting my performance marks inside my HTML, I switched to the “Performance” tab, made sure I selected a “Fast 3G” network and “4x slowdown” for the CPU

      It's worth checking the profile on sitespeed.io to see how this compares

    2. Since I only wanted to see the CSS coverage, I used the filter “.css” and what I could see was that 92% of the CSS I was loading was not used. (Unused bytes will change in real-time when you start interacting with the page):

      Is this exposed in sitespeed?

    3. Structure of my page after adding the performance marks

      Breaking out the CSS, JS and head tags, to time them all independently

    4. So, I decided to use some custom metrics using the Performance API to get a rough idea of what was time-consuming on the page I was auditing.

      So, using the performance API to translate it into something meaningful for clients

    1. “the tragedy of the horizon.”

      New term to me. It's good.

    2. to avoid mass death.

      This is as good a climate slogan as any, right now.

  6. Apr 2020
    1. For example, the GUI tool for PostgreSQL administration, PGAdmin 3, is used by many people.  (I’m an old-school Unix guy, and thus prefer the textual “psql” client.)  I’ve discovered over the years that while PGAdmin might be a useful and friendly way to manage your databases, it also automatically uses double quotes when creating tables.  This means that if you create a table with PGAdmin, you might find yourself struggling to find or query it afterwards.

      Oh my god. I wish I knew this ten years ago.

    1. Averylarge data centre may consume 30GWh of power in a year, costing its operator around £3,000,000 for electricityalone. A handful of sites in the UK consume even more than this although the majority of sites consume far less. The total power demand of the UK data centre sector is between 2-3TWh per year2. Energy is usually the largest single element of operating costs for data centres, varying from 25-60%.

      This is about one percent of UK electricity usage. And this seems to discount smaller datacentres, which make up a much larger use of power is the datacentre dynamics piece from john booth is anything to go by.


    1. In this case, establishing a direct connection between thepeers seems to be impossible, and the only remaining op-tion is to use a server with a public IP address to proxythe communication between the peers, e.g. using the TURNprotocol [18]

      This is just like our experiences with video again then. Back to TURN.

    2. For ephemeral updates PushPin uses an additional mes-saging channel, adjacent to the CRDT, which ties arbitrarymessages to a device and user context. The current implemen-tation is rudimentary: ephemeral data is not associated witha particular CRDT state and is distributed only over directP2P connections. Nevertheless, it enables shared contextualawareness in the user experience of PushPin, providing afeeling of presence when other users are online or collabo-rating

      So THAT's how they do the presence like mouse pointers and the rest.

      If there's already a channel here, presumably you could do video too, if there was a central server

    3. Each URL also includes acontentTypeparameter, whichindicates how that document should be rendered in the userinterface. This parameter is part of the URL, not the docu-ment content, because the same document content may berendered differently in different contexts. For example, Push-Pin could be extended to support flashcards for languagelearning. In one context, the document containing the data-base of flashcards could be rendered as a list of entries, whilein another context it might be rendered as a quiz interface,presenting one side of one flashcard at a time.

      I had no idea that there was a separate 'viewer' concept. Neat

    1. People make light of the idea that digital should be the most basic of Maslow’s hierarchy of needs — over food, water, shelter, and warmth — but there is evidence that people do, to an extent, prioritise connectivity over food and comfort. Some refugees, for instance, are known to have asked for Wi-Fi or charging services ahead of food or water on arrival in a new country.


    1. Globally, Amazon has 86 solar and wind projects that have the capacity to generate over 2,300 MW and deliver more than 6.3 million MWh of energy annually—enough to power more than 580,000 U.S. homes.

      The use of "has the capacity to" here is misleading.

      Capacity factors for renewables range betweeen 15 and 40, so this makes it sound like amazon has paid for and used 6.3 MWh of power when it's likely to be a fraction of this.

      By comparison, Microsoft, report on how much power they did use each year, and how much of that was renewable. Last year they reported using around 7m MWh of power, and buying almost this figure in renewable energy.

  7. www.gitops.tech www.gitops.tech
    1. Additionally the image registry can be monitored to find new versions of images to deploy.

      Ahh, so this is how you might make it possible for a developer to just push changes, and not have to have access to the environment repository.

    1. For now, suffice it to say that Tailscale uses several very advanced techniques, based on the Internet STUN and ICE standards, to make these connections work even though you wouldn’t think it should be possible. This avoids the need for firewall configurations or any public-facing open ports, and thus greatly reduces the potential for human error.

      I wonder how this relate to VOIP and so on?

    2. Here’s what happens: Each node generates a random public/private keypair for itself, and associates the public key with its identity (see login, below). The node contacts the coordination server and leaves its public key and a note about where that node can currently be found, and what domain it’s in. The node downloads a list of public keys and addresses in its domain, which have been left on the coordination server by other nodes. The node configures its WireGuard instance with the appropriate set of public keys.

      So it's a little bit like a private DNS server?

    3. Sadly, developers have stopped building peer-to-peer apps because the modern Internet’s architecture has evolved, almost by accident, entirely into this kind of hub-and-spoke design, usually with the major cloud providers in the center charging rent.

      This is such a quote.

    1. A breadcrumb in this case is a single pixel that you can place in a precise location on a webpage. Placing a breadcrumb could be as simple as Option + click. While navigating the web, you could leave breadcrumbs on different pages you find interesting over the course of a browsing session. When you're done, that sequential "trail of breadcrumbs" would be saved. You could then jump back into the trail and navigate "forward" and "backward" through the things you found interesting in that browsing session. Or share the trail with a friend, and they could step through your spatial path of navigating the web.

      This isn't a million miles away from how hypothesis allows you to annotate specific sections, although much more lightweight.

      It's plausible to show a set of thumbnails of a pages with the highlights ... highlighted for others to see.

    2. When you run a standup at a technology company, you typically go around a circle and each person gives their daily update. But in Zoom, there is no circle. You get confused about who is next. Two people start speaking at the same time. It's awkward and confusing. Eventually, you realize that one person, probably the manager, just has to dictate who goes next, at the risk of seeming bossy.

      Just having a line from left to right would be an improvement, or some consistent, implied order

    1. Since emission reduction projects registered under crediting programmes to date have been mostly developed in the context of cost-saving, rather than ambition-raising mechanisms, we understand that there are very few, if any, examples of existing credited projects that represent those high-hanging fruits, and which could be considered truly additional in the context of the Paris Agreement. Given the difficulty in objectively determining additionality in line with this definition, we consider that only a niche and ever reducing number of activities could count for this, and that this does therefore not represent a viable option for rapidly increasing demand volume of the market.

      This is a really important point, that the current "offsets" framing undermines.

    2. A climate responsibility approach needs to first and foremost incentivise and facilitate the reduction of one’s own emissions.

      I can't help thinking they should have lead with this

    3. Emissions from project-specific activities, such as project-related travel, are attributed as cost items to their respective project cost lines.

      Project level carbon budgeting?

    4. We recognise that some of the activities with the highest transformation potential – and therefore with high suitability for supporting the objectives of the Paris Agreement – may be at early stages of development and/or may carry a risk of not delivering attributable emission reductions.

      This addresses some of the issues around needing to get a definite amount of emissions drawn down, when the science makes this very very hard to measure

    1. The High-Level Commission on Carbon Prices surveyed the available scientific literature, concluding that the explicit carbon-price level consistent with the Paris Agreement temperature objectives is at least US$40–80/tCO2 by 2020, provided that a supportive policy environment is in place (High-Level Commission on Carbon Prices, 2017). Informed by this report and allowing for its uncertainties, NewClimate Institute has imposed a price level of EUR 100/tCO2e for the 2014-2019 period. This is also in-line with the central estimate of climate change avoidance costs over the period to 2030 used in the European Commission’s 2019 Handbook on the External Costs of Transport (European Commission, 2019).

      Holy biscuits. They're not fucking about. Microsoft just increased their price for carbon to 14 USD per tonne, by comparison.

    2. This methodology for the estimation of GHG emissions includes the estimated equivalent climate impact of non-carbon climate forcers from aviation, such as condensation trails, ice clouds and ozone generated by nitrogen oxides and results in emission estimates approximately three times greater than if calculating only direct CO2 emissions (Atmosfair, 2016).

      Wow, they use the updated science

    1. For the moment, the main take-away is that there is a goodargument that registration fees should ​not​​be set at $0, even this year. Rather, organizersshould look at their existing budgets, and rework them by eliminating the costs associated withthe physical event. Virtual conferences that do choose to set their prices low should be carefulnot to encourage an expectation that other virtual conferences (or future instances of this one)will always be free or cheap. (For example, one of the suggestions heard by the ASPLOSorganizers was “keep it free;” obviously, this may not be financially sustainable.)

      I'm really glad this is mentioned, as it provides a chance to talk about actually paying presenters for their time, increasing the likelihood of new voices speaking at conferences.

  8. Mar 2020
    1. He mentioned a few examples where the AI fallacy is already playing out. One is the idea from the national statistician that we might not need to keep doing a census, because there will be lots of data from other sources. Neil pointed out that we actually need more classical statistics than ever, to verify all this machine learning and data. He calls this the “Big Data Paradox” - that as we measure more about society, we understand less. We need to be able to sanity check our large complex systems - the census is still valuable.

      The census as a calibration tool

    2. In the section on technical debt, the mythical man month, etc, I was amused to note that Neil called out Amazon's fabled "two pizza team" as American cultural imperialism. The problem arising from the separation of concerns and specialisation of teams is that no one is concerned with the whole system.

      This is such a good quote and helps explain so much.

    1. GENEVA (Reuters) - European countries need to invest to prepare their transport infrastructure for the impacts of climate change or face hundreds of millions of dollars in repair costs, a U.N. regional commission said in a study it says is the first of its kind.

      This is for transport. What about networks?

  9. www.fairphone.com www.fairphone.com
    1. GWPkg CO2e43.8535.98-1.115.983.00100.0%82.1%-2.5%13.6%6.8%

      These figures here, for a smart phone from 2016 put the use phase at around 6kg of CO2 over a 3 year life cycle.

      That's fairly close to the Shift Project figures in their Lean ICT report.

    2. The following use pattern is applied for the Fairphone 2:

      This assumes similar energy use figures to the Shift project of 5.9kwH per year

    1. The quantification of this unit impact is done in kWh/byte. Three contributions are considered:The electricity consumption associated with using the terminal on which the action is performed;The electricity consumption generated by the activity of the data centers involved in transferring the data;The electricity consumption generated by the activity of the other network infrastructures during the transfer of the data.

      Awright. So, reading this makes me think all the numbers from the 1byte model that are being cited appear to be concerned with the use phase, not the actual production phase.

  10. Feb 2020
    1. Fundamental requirement:Climate-neutral datastorageand transfer

      I'm curious about how this is measured and managed.

    1. Data centres and telecommunications will need to become more energy efficient, reuse waste energy,anduse more renewable energy sources. They can and should become climate neutral by 2030

      For maddie

    2. Destination Earth, initiativeto develop a high precision digital model of Earth (a “Digital Twin of the Earth”) that would improve Europe’s environmental prediction and crisis management capabilities (Timing: from 2021

      Good heavens, a digital twin of Earth

    3. A circular electronics initiative, mobilising existing and new instruments in line with the policy framework for sustainable products of the forthcoming circular economy action plan

      I wonder if the restart project gang have seen this

    1. Climate change poses an unprecedented threat to humanity in the 21st century. In the period up to 2030, an estimated $3.5 trillion is required for developing countries to implement the Paris climate pledges to prevent potentially catastrophic and irreversible effects of climate change.

      Ah, that's where it comes from

    1. Methodology and sources The analysis of the carbon intensity of streaming video presented in this piece is based on a range of sources and assumptions, calculated for 2019 or the latest year possible. Bitrate: global weighted average calculated based on subscriptions by country and average country-level data streaming rates from Netflix in 2019; resolution-specific bitrates from Netflix.Data centres: low estimate based on Netflix reported direct and indirect electricity consumption in 2019, viewing statistics and global weighted average bitrate (above); high estimate based on 2019 cloud data centre IP traffic from Cisco and energy use estimates for cloud and hyperscale from IEA.Data transmission networks: calculations based on Aslan et al. (2017), Schien & Priest (2014), Schien et al. (2015), and Andrae & Edler (2015), and weighted based on Netflix viewing data by devices. Devices: smartphones and tablets: calculations based on Urban et al. (2014) and Urban et al. (2019), iPhone 11 specifications (power consumption and battery capacity), and iPad 10.2 specifications; laptops: Urban et al. (2019); televisions: Urban et al. (2019) and Park et al. (2016), and weighted based on Netflix viewing data by devices. Carbon intensity of electricity: based on IEA country-level and global data, and 2030 scenario projections.

      Let's roll these into the new model

    1. Aristotle’s view of the process: “We are what we repeatedly do. Excellence, then, is not an act but a habit.”

      This is a really handy quote. totally borrowing it.

    2. Buying carbon offsets might still have greater impact in the short run, but you can’t see them, so their purchase is less likely to be contagious.

      I've never thought of them in terms of 'contagion' like this. interesting

    1. Today the average carbon intensity of electricity generated is 475 gCO2/kWh, a 10% improvement on the intensity from 2010.

      There is more recent data, but it's not for the whole world. Should come out around March 2020.

    1. It is relevant to estimate how much data is generated by - and associated electric power used - normal behavior like video streaming several hours every day.

      If we assume video streaming is 'normal behaviour' like web surfing and working online, then these numbers are probably the safest to use for now.


    1. TechnologyIntensity [kgCO2eq/MWh]solar0.00410geothermal0.00664wind0.141nuclear10.3hydro16.2biomass50.9gas583unknown927oil1033coal1167

      The proportions of each are going affect this.

      What would a global average (mean) figure be for renewable energy (i.e. not including nuclear)?

    2. Fig. 1. The 28 areas considered in this case study, and the power flows between them for the first hour of January 1, 2017. The width of the arrows is proportional to the magnitude of the flow on each line. Power flows to and from neighboring countries, e.g. Switzerland, are included when available, and these areas are shown in gray. The cascade of power flows from German wind and Polish coal are highlighted with blue and brown arrows, respectively.

      I had no idea Germany sold so much power, net to other countries. Always assumed it bought loads of France's power

    1. So energy consumption in the oil and gas sector, into which Google Cloud is selling its cost-reducing services, is four orders of magnitude larger than Google’s data center decarbonization efforts. The harm that Google Cloud will do to the planet, if it reduces underlying costs of this industry by even a small percent, completely dwarfs the data center decarbonization work.

      This is the first time I've seen numbers putting these into perspective. I wonder how these compare for Amazon and M$ ?

    1. Therefore, we estimate an advertising share of 50% for the traffic class web, email, and data in 2016, with an uncertainty range of [25%–75%]. The share is the same for both mobile and fixed traffic.

      Holy biscuits, HALF OF WEB TRAFFIC as ads?

    2. According to a Solarwinds company 2018 study, the average load time for the top 50 websites was 9.46 s with trackers and 2.69 s without.

      Trackers have a 4x impact on performance compared to not having them

    3. The Internet's share of the global electricity consumption was 10% in 2014 (Mills, 2013): As a reference, the entire global residential space heating in 2014 consumed the same amount (International Energy Agency, 2017a).

      Heating ALL the homes in all the world is about the same as the internet's carbon footprint according to this paper

  11. Jan 2020
    1. How to Build Data Visualizations in Excel

      Does this exist for Google Spreadsheets or Libreoffice?

    1. . In particular,on the host, we propose a first configuration of a software-defined power meter that builds on a new CPU power modelthat accounts for common power-aware features of multi-core processors to deliver accurate power estimations at thegranularity of a software process. In the VM, we introduce asecond configuration of a software-defined power meter thatconnects to the host configuration in order to distribute thepower consumption of VM instances between the hosted ap-plications. The proposed configuration can even be extendedto consider distributed power monitoring scenarios involvingapplication components spread across several host machines.

      So, this would be the app level metering you would want.

    2. WhileBITWATTSis a modular framework that can accom-modate different power models (includingrunning averagepower limit(RAPL) probes and power meters), we propose aprocess-level power model, which is application-agnostic andaccounts for virtualization—i.e., for emulated cores withina VM—and for the power-aware extensions of modern pro-cessors, notably hardware threading anddynamic voltageand frequency scaling(DVFS).

      Ah, so this is the key difference and the reason for the Smartwatt formula business

    1. the politics of the armed lifeboat

      I expect to use, and hear this term a lot in 2020, as people realise the implications of changing climates on how we live.

    1. For Apple, this may be a feature rather than a bug: Documents obtained by Motherboard in 2017 revealed that the company requires its recycling partners to shred iPhones and MacBooks so that their components cannot be reused, further reducing the value recyclers can get out.

      not great for a parts market

    2. So, Apple has started collecting that scrap, melting it down and forming new hunks of aluminum that can be used to carve more gadget husks.

      Wait, this is the recycling?

    1. In February 2018, she published a report — titled Mission-Oriented Research & Innovation in the European Union — that defined five criteria missions should obey: they must be bold and inspire citizens; be ambitious and risky; have a clear target and deadline (you have to be able to unambiguously answer whether the mission was accomplished to deadline or not, Mazzucato says); be cross-disciplinary and cross-sectorial (eradicating cancer, for example, would require innovation in healthcare, nutrition, artificial intelligence and pharmaceuticals); and allow for experimentation and multiple attempts at a solution, rather than be micromanaged top-down by a government.
      1. bold and inspiring to citizens
      2. ambitious and risky
      3. clear target and deadline
      4. cross disciplinary and cross sectortial
      5. allow for multiple approaches and experiments
    2. Mazzucato traced the provenance of every technology that made the iPhone. The HTTP protocol, of course, had been developed by British scientist Tim Berners-Lee and implemented on the computers at CERN, in Geneva. The internet began as a network of computers called Arpanet, funded by the US Department of Defense (DoD) in the 60s to solve the problem of satellite communication. The DoD was also behind the development of GPS during the 70s, initially to determine the location of military equipment. The hard disk drive, microprocessors, memory chips and LCD display had also been funded by the DoD. Siri was the outcome of a Stanford Research Institute project to develop a virtual assistant for military staff, commissioned by the Defense Advanced Research Projects Agency (DARPA). The touchscreen was the result of graduate research at the University of Delaware, funded by the National Science Foundation and the CIA.

      This paragraph.

  12. Dec 2019
    1. Just this month it was announced that over half of the power plants operated by China’s Big Five state-owned utilities are running at a loss. The government has plans for up to one third of them to shut by 2021, removing 15% of the country’s coal capacity.


    2. No matter that there is no single robust methodology: internalizing Carney’s three types of climate risk – physical, liability and transition – will inevitably work through to large-scale asset disposals and new reinvestments; it will without doubt affect the cost of capital of high-carbon businesses.

      This taxonomy of risk sounds useful to talk about.

    3. No single “sneeze” will wipe out fossil fuel use across energy and transport; It will occur sector by sector, country by country. Over the past six years, LED light-bulbs have gone from less than 5% global market share to over 40%; coal power in the U.K. from 40% to a couple of percent; plug-in vehicles in Norway from less than 5% to over 50%. In each case, there was a slow start, an agonizing wait, and then the sneeze. Bless you!

      Good reference for explaining a punctured equilibrium

    4. The average capacity factor of the world’s hydro plants is 42%; gas peaking plants 15%. Even so-called baseload coal plants run on average only 54% of the time. If technology is cheap, and demand or supply are intermittent, we overbuild. Wind and solar are no different.

      Cripes. I didn't know that even baseload coal runs so infrequently. From the public discourse you'd assume it was 24/7

    5. economic drag caused by resource waste

      I've never heard of this, but I assume it's basically shifting from saying "we'd grow faster if only we had another two power stations" to "we'd grow faster if we used energy create the effect of having two more power stations of useful energy"


      Seminal paper on how we navigate online with search

    1. Starting with a small cohort, we’ll refine the patterns and practices that sustain learning. We will see case studies of how the organizing model is made locally relevant. Globally, we will be a large peer learning community. And we’ll also have smaller, local peer learning communities connected and experimenting in a place.

      This is very applicable in CAT. Get a small core to make the ground game.

      lol… ground CATs

    2. For our community that teaches the web, we understand ourselves as participating in i) a global network and ii) a local context.

      What is the narrative for CAT like this?

    1. The general argument I see in this paper is largely:

      • most companies end up reporting CO2 emissions from energy use misleadingly, because the standards allow them to claim credit using a financial instrument that's not proven to be very effective.
      • worse, reporting like this causes double counting elsewhere, as other companies rely on an average figure that doesn't take into account these RECS
    2. However, to be accurate and relevant, GHG inventories must reflect the emissions caused by the reporting entity. The fundamental issue with the contractual method is that it does not represent any causal relationship between the reporting entity and the emissions reported.

      OK, so this is little bit like, like how a lagging indicator, reported at a low enough resolution won't provide enough of a signal to actually drive changes in behaviour for an org or team

    3. The distinction between attributional and consequential GHG accounting is well understood in some fields, such as life cycle assessment (LCA), but much less so in the area of corporate GHG accounting, which has traditionally used attributional accounting (Brander et al., 2015).

      Attribution and consequential accounting. Need to research these phrases

    4. However, in reality, contractual emission factors represent a market failure, because the implied goal (actual reduction of emissions to the atmosphere) is in fact not delivered by contractual emission factors (as discussed in Section 2), and therefore the approach has only the appearance, and not the substance, of a market-based solution.


    5. Further, for Company A's use of contractual emission factors not to lead to double-counting of the claimed renewable attributes, the method requires all other reporting entities to also apply a ‘residual grid mix’ emission factor. As this emission factor would be higher than the grid average, due to removing some renewable generation from the calculation, Company B's performance is again made to look worse, although its actual contribution to reducing emissions is greater than Company A's.

      So if they claiming the market based figures here, the avg figures on the grid would need to look dirtier, for everyone else, but it doesn't.

    6. Electricity generation accounts for approximately 25% of global greenhouse gas (GHG) emissions, with more than two-thirds of this electricity consumed by commercial or industrial users

      Two thirds comes from industry - useful figure when talking smart meters in the home

    7. This issue is highly topical as recently published reporting guidance from the GHG Protocol (WRI, 2015) has endorsed the market-based approach, while the forthcoming update of ISO 14064-1 for corporate GHG inventories provides an opportunity to establish a more robust approach.

      This was last year. I wonder if the guidance has been watered down, or made stronger.

      Recent reporting suggests not.

    8. The GHG Protocol's Scope 2 Guidance, published in 2015, requires that companies use both the locational grid average method and the market-based method to report scope 2 emissions (i.e. dual reporting). However, the guidance also allows companies to choose a single method for meeting their reduction targets and for reporting their supply chain emissions (WRI, 2015). The same guidance has been adopted by CDP, formerly the Carbon Disclosure Project (CDP, 2016a).

      This doesn't seem clear. This means that for reporting, they need both - if this is the case then companies that publish only one are basically not following the guidance properly.

    9. Moreover, in many countries the amount of renewable generation is increasing due to the other drivers, such as government subsidies (IEA, 2016b), and therefore the point at which additionality might be achieved (i.e. beyond Q1) is continually advancing further beyond the reach of voluntary market demand for contractual emission factors.

      So, they're changing anyway, and the RECs don't have much impact.

    10. ET Index Research data, which includes 2000 of the world's largest listed companies, shows 97 companies using the market-based accounting method to report lower emissions, equating to 22.2 million tCO2e/yr.2 This approximates to ~ 1% of globally available renewable electricity generation in 2015,3 and therefore demand for contractual emission factors would need to increase a hundred-fold to reach the existing supply threshold for renewable attributes (which is continually increasing anyway), and only once above that threshold would demand cause a fractional increase in renewable generation.

      We're massively under reporting corporate CO2 emissions from energy use, and the majority of companies are basically hiding the real figures

    1. New School’s Digital Equity Lab

      I had no idea this existed! Greta sounds like a cool cat

    2. Chris Adams, a web designer and climate activist in Berlin, tells me he thinks a green internet must be free of advertising. “Ninety percent of a web page being ads requires servers, and those servers are taking electricity, and that electricity is generated by burning coal,” he says. Adams has written that the European site for USA Today is a model of efficiency. It removed all of its tracking scripts and ads to be compliant with recent General Data Protection Regulation legislation in the European Union. The site size immediately shrank from 5 megabytes to 500 kilobytes, but it still basically looks the same—there are just no ads.

      I don't think the internet should be free of all advertising, and I have never said this. I'm not sure it CAN be.

      That said, not all advertising is equal - some is really scummy, invasive stuff that can get in the sea, but to say I think this feels like a reach.

    1. Describing the move as an industry first, Repsol said it wanted to lead a wider transition to renewable energy, in line with the goals of the 2015 Paris Agreement to avert catastrophic climate change.

      Oil companies are getting out of oil, and still people frame access to energy in terms of getting cheap out the ground oil.

    2. Giving extra impetus to its new goals, Repsol will link at least 40% of managers’ long-term variable pay to its emissions reduction targets.


    1. Build on narrative that integrates the two approaches, probably building oncircularity (be part of the coming standard) and efficiency of services (e.g. theexample of interface not knowing their impact in services besides havingenvironmental product declarations for all their products).

      Key thing. Inward sustainability focus misses the big picture


    1. In short, Nordhaus, who is mentioned both in my 2007 and 2012 pieces, tells us not to worry too much about climate change. It will be cheaper to adapt to it than to prevent it or slow it down.

      Oh jeez, so THIS is where that adaptation vs mitigation meme came from. In the context of what research gets funded, why he got the Nobel prize is making more sense now.

    1. Basically, all services in public administration will moveinto the cloud, which means that a huge amount of data will be digitalised. This has huge implications forthe security of the data (which can quickly translate into national security matters). Most of the services are now provided by companies outside Europe, which raises questions aboutdata sovereignty, ownership, transpar-ency, etc. Also, as recently proposed by Germany (GAIA-X project), the European Union should re-gard the cloud as a critical infrastructure which should be developed within Europe.



    1. Almost a third of UK boards (32%) feel little or no responsibility for the climate crisis, according to a global survey of 640 chairs and non-executive directors by recruitment firm Harvey Nash and London Business School’s Leadership Institute.


  13. Nov 2019
    1. The emissions from onereturn ticket from London to New York are roughly equivalent to that of heating a typical home in the EU for a whole year (European Commission, 2019).

      Holy balls. I've never seen it compared in those terms before.

    1. Those instances of decoupling that can be observed (like the UK or Germany in past years) result mainly from deindustrialisation and the outsourcing of energy-intensive industrial production to other countries.

      I haven't found a good rebuttal to this. I'd love to find one.

    1. energyusage.evaluate(exp, 10)

      So, you work out the figures by passing the function and the params for it.

    1. The questions within this Guide are organized into three relevant impact areasto assist with this evaluation: EnvironmentalPractices and Policiesof the Cloud Service Providers; Data Center Facility Management and Equipment; and Data Center Power Sources.

      So, basically:

      • policy
      • how they provision capacity
      • how they source power
    2. In the Fallof 2017, GECsponsored an Arizona State University(ASU) graduate capstone project that foundthatCloud Service Providers’publicly available sustainability data was limited,and that thesustainabilitymetrics and terminology used by those providers wereinconsistent and confusing. The ASU research findings assisted in thedevelopment of this Guide.

      So, basically every group say it's a trainwreck right now, transparency-wise.

    3. We meet our mission by supporting institutional purchasers in leveraging their purchasing power forsustainableproductsand servicestoadvance the market for those products.

      Procurement as a lever, then.

    1. easyJet’s aircraft carbon emissions in the 2018 financial year were 7.6 million tonnes, compared to 7.1 million tonnes in the 2017 financial year. easyJet’s calculation of emissions is based on fuel burn measurement, which complies with the EU’s Emissions Trading System requirements.

      7.1m in a year in 2018. That's about 3 and a half Googles.

    1. More figures extracted from the EURECA project, 2018. 40% of servers in public sector data centres were over 5 years old. These performed 7% of compute but used 66% of power.

      Wow, that's a stat.

    2. UK Government departments are required to report energy usage and explain how they are reducing their respective carbon footprints and this obligation to understand and report scope 3 emissions is likely to become more widespread, given UK government commitments to climate change targets and increasing observance of UN Sustainable Development Goals (SDGs). It is also a requirement of the Science Based Targets methodology.

      Ah, it's a requirement of the Science Based Targets methodology. So any company signing this will need to ask the questions.


    1. or

      So OR is important. If you're an EU citizen and self employed and working, you have the right to stay.

      Even if you're an EU citizen, and not working, as long as you have enough money to support yourself, you have the right to stay.

    2. The right of residence provided for in paragraph 1 shall extend to family members who are not nationals of a Member State, accompanying or joining the Union citizen in the host Member State, provided that such Union citizen satisfies the conditions referred to in paragraph 1(a), (b) or (c).

      OK, so the folks at the Auslandeborde basically lied to our faces. Jeez, all this stress the last few weeks.

    1. Amazon's carbon intensity metric, measured as grams carbon dioxide equivalent (CO2e) per dollar of Gross Merchandise Sales (GMS), equals 128.9 g CO2e per dollar (USD).

      My guess is that AWS's emissions per unit of spend will be lower. Reporting only market based emissions makes it hard to tell how much energy is really in use, but for e-commerce companies, more than half the CO2 comes from delivery.

    1. For example, spending on parcel delivery by third-party carriers is mapped to the couriers and messengers sector (i.e., 224 grams of CO₂e per dollar) and shipping boxes are mapped to paperboard container manufacturing (i.e., 807 grams of CO₂e per dollar).

      This makes me think there's a figure for the use of AWS itself too.

    1. However, the table and figure below both use 2012 datapublished in 2014for all three different approaches to reporting emissions.

      Basically, embedded emissions are 10-20% more than the UK environmental account ones

    2. Embedded emissions3”, published by the Department for Environment, Food and Rural Affairs (Defra)–these measureemissions on a “consumption” basis, and take account of the emissions embedded within the manufactured goods and services which the UK imports and exports.

      AHhh! this sounds like the thing that would be ideal if it was on a per sector basis.

  14. Oct 2019
    1. If the current de-velopments determined by Borderstep continue, the energy consumption of data centers will double by 2030 compared to today.

      So, doubling from today assumes around 700 kWh in 2030

    2. While Shehabi et al. assume that maximum power consumptionis constant[4], the Borderstep model assumes an increase in maximum power consumption due to a significant increase in the average amount of RAMs and multiprocessor systems.

      Ah! That makes sense - a big load out in a server, with loads of cores and RAM is deffo gonna make it need more juice.

    3. According to a Borderstep Instituteestimate, worldwide energy consumption of server data centers increased by about 30% to 287 billion kWh between 2010 and 2015 [9]. This in-crease accelerated once again in the last two years. A current TEMPRO project estimate concludes that between 2015 and 2017 theenergy consumption of data centers worldwide in-creased by approx. 20% to 350 billion kWh

      By comparison, the IEA figures generally say energy use has stayed around 200 kWh in 2020.

    4. The analyses and results presented in this article were pro-duced as part of the project TEMPRO –"Total Energy Man-agement for professional data centers." TEMPRO is supported through the 6th Energy Research Programme of the German FederalGovernment.

      Are there similar studies in other countries?



    1. Online video viewing, which represents 60% of the world’s data traffic, generated more than 300 MtCO2e during 2018, i.e. a carbon footprint comparable to the annual emissions of Spain

      This here is the stat I link in this talk:


      As far as I can tell, it's using the 1bit model from the shift project. Other methodologies give different examples, but it's at least documented.

    1. “The horror of climate change isn’t in the intrinsic violence of hurricanes or heat waves, but in the ways societies choose to deal with and prepare for them.”

      This quote. wow.

    1. Taking into account the global electricity mix, the share of greenhouse gas (GHG)emissions attributable to the Digital era would thereforeincrease from 2.5% in 2013 to 4% in 2020(2.1 Gt)

      This is the figure for Sal.

    1. ⁴ Amazon reports emissions according to the GHG Protocol’s “market-based” method, which accounts for renewable energy Amazon purchases to support its operations.

      Market based then.

    1. MORFBEATS Where art meets sound

      These are the horror music devices

  15. Sep 2019

      ~1.2m with market based adjustments


      4.9 without market based adjustments

    3. Our carbon footprint

      Before offsetting, but after market based emissions, it's about 1.2m tonnes of CO2

    1. It derives emission factors from contractual instruments, which include any type of contract between two parties for the sale and purchase of energy bundled with attributes about the energy generation, or for unbundled attribute claims. Markets differ as to what contractual instruments are commonly available or used by companies to purchase energy or claim specific attributes about it, but they can include energy attribute certificates (RECs, GOs, etc.), direct contracts (for both low-carbon, renewable, or fossil fuel generation), supplier-specific emission rates, and other default emission factors representing the untracked or unclaimed energy and emissions (termed the “residual mix”) if a company does not have other contractual information that meets the Scope 2 Quality Criteria

      Blimey, this could be much more than using RECs.

    2. Not having contractual data for every site will not cause noncompliance with the GHG Protocol Corporate Standardand Scope 2 Guidance

      This is how amazon can still pass the audit. It's not strictly necessary, but it's a bit sneak to not do it. Amazon, after the using instruments to make it look like it uses less electricity, appears have had used a similar amount to google did, before, google applied their own market based methods.

    3. Companies with any operations in markets providing product or supplier-specific data in the form of contractual instruments shall report scope 2 emissions in two ways and label each result according to the method: one based on the location-based method, and one based on the market-based method. This is also termed “dual reporting.”

      So, here's the thing. You really should report both figures, as it gives an idea of the impact of the different instruments in use, like purchasing renewable energy credits, and so on.

      If you report location based, it at least shows how much electricity you might be using, before the market based approach of buying RECs or similar brings the figure right down.

      If you don't know what market based instrument a company is using, you have no idea if they're using some skanly RECs bought in some rando part of the world, to give artificially low figures.

    1. there will also be a smaller event in 2019 as part of the Dutch Design Week programme, a Service Design Challenge for social design based on a real-life challenge on Saturday October 26.

      Okay, this is worth keeping an eye on

    2. Many large organisations have embedded service design standards and methods into their corporate procedures — such as Transport for London — reflecting a significant increase in their internal capacity.

      Oh, I did NOT know this. Cool

    1. In a perfect CE the residence time of a unit of material in thetechnosphere would be indefinite.

      technosphere presumably means, in use?

    2. At the same time, due to the higher specific GHG intensity of alu-minum, the GHG emissions of vehicle manufacturing would increase toenable life cycle savings in future years

      So this is what the chart is showing. lightweighting makes the production phase more carbon intensive, to make the use phase lower

    3. Monitoring and managing the de-velopment of in-use stock is thus central to the transition to a morecircular socioeconomic metabolism, which is why a separate stock ca-tegory is part of the dashboard.

      so it's like limited WIP, but for steel


    1. When we purchase solar and wind energy, we use the GHG Protocol’s market-based method to demonstrate how renewable energy purchases—from Amazon Wind Farm Texas, for example—reduce our consumption of grid electricity.9

      this is new to me.

    2. Amazon Devices: We developed a specialized carbon footprint model to address the complexity of the manufacturing, use, and end-of-life of Amazon devices, including Echo devices, Kindle e-readers, Fire Tablet, Fire TV, Ring, Blink, and all others. This starts at the component level—including where components are sourced and how they are manufactured into products—and extends to customers’ use of the product and the eventual product end-of-life.

      Okay, and these here sound like bespoke models like you might make using AMEE

    3. Financial: We combine data on Amazon’s spending with industry-specific, dollar-based emissions factors (e.g., a standard amount of carbon dioxide pollution associated with every dollar of spending on a particular activity). We use this model to capture carbon emissions from activities like the construction of Amazon buildings, the manufacturing of Amazon’s Private Brands products, equipment used in our warehouses, office supplies, and other purchased goods and services.

      And this here sounds like they're using an input/output based approach like industrial ecologists are doing.

    4. Our approach to quantifying our carbon footprint reflects the complexity of our business. Our team of researchers and scientists have combined cutting-edge life cycle assessment (LCA) science and Amazon Web Services (AWS) big data technology to develop a robust software solution that processes billions of operational and financial records from Amazon’s operations across the world to calculate our carbon footprint. The software estimates carbon emissions for all activities within our system boundary using a dollar-based environmental assessment model, then enhances the accuracy of carbon-intensive activities with detailed, process-based LCA models.

      This sounds a lot like a top down approach as outlined by systems like Ecocost or Trucost, then supplemented with bottom up modelling of specific processes

    1. The Panoply, a Code of Conduct that encompasses environmental and ethical standards.

      This links results in a 404 file not found message right now.

    2. “It’s pretty easy to decarbonise,” says Adams. “We’re just not doing it fast enough right now.”

      Oh wow, without the rest of the sentence, this sounds really out of context.

      It's HARD to decarbonise in general (steel, shipping, aviation for example are all really tough) but tech by comparison has it easy.

      Cloud and IT services are much simpler to decarbonise than the other sectors, partly because our experience of a site doesn't really change depending whether it's running on green power or not.

    1. While in some parts of the world, you can buy power from a company that only sells renewable power, there are other parts of the world where the grid is regulated to the point that you don’t even have choice of supplier – if they burn coal, and you need to run a server in that part of the world, there’s no other option. Elsewhere, there may be places where company A sells renewable power, but also sells the fact that energy has been generated with renewable power separately, so that another company B, can buy this credit, and then say they use green power, even if the supplier they use is relying on fossil fuels to generate electricity for their servers .
    1. I was given new language that helped described my own personal experiences of how care services (for example) are often designed for the individual when in fact they need to be designed for the system, unit or in systems coaching language, ‘the third entity.’ The Third Entity is that which lives outside the individuals in a particular relationship system whether a work team, family, community or society at large.

      the third entitiy is the outer ring

    1. Purchase green power certificates. In July 2017, China launched a pilot program that permits voluntary trade of green power certificates from solar and wind power. Each certificate represents 1 MGh of electricity. Buying green power certificates allows companies to claim environmental benefits associated with renewable energy generation, even if electricity from a renewable power plant does not feed directly into a data center facility.

      Oh, wow, they do RECS too now? I wonder if they publish them too

    2. 196.96

      this forecast is almost the same as the IEA's earlier report for global demand for energy from datacentres

    1. 194

      This basically says that efficiency gains, mainly from cloud, are holding back absolute growth in energy use.

  16. Aug 2019
    1. Much of this will be lost if the quiet time and ability to think independently is taken away, and if, in particular, the discussion portion is lost.

      quiet time is disrupted when people are writing into the same mutually visible section of a google doc, or mural for example.

    2. act and influence

      There's a deliberate decision not to talk about the things you have no control over right now

    3. Everyone can vote all at once. Are the consequences positive, negative, or both?

      Okay, this is useful prompt, rather than just saying "what are the good things, and are the bad things?"

    4. A copy of your company’s vision, mission and values

      This is good to mention, based on an earlier run through we did.


    1. This is where the insight from Web IDL comes in. When you squint at it, Web IDL kind of looks like an IR.

      wow, this is much more ambitious than I first realised

    1. The energy consumption of ICT equipment in times of low utilization is still a big issue, even if the 2019 ecodesign regulation on servers and storage products will lower the idle consumption.

      ecodesign regulation? what is the equiv outside of yurp?

    2. Server utilization has long been an important issue, as many servers are often idle and although they do not do any work, they still consume power

      what is typical utilisation here?

    3. For 5G mobile radio it is already clear that the energy consumption per GB will be considerably lower than in previous generations,

      Why would this be?


    1. While this is absolutely true, if you care about performance, you should be running your own content from a CDN already. With the price of modern hosting solutions being what they are (this site is fronted by Cloudflare which is free), there’s very little excuse for not serving your own assets from one.

      You can run a CDN without needing to have a different origin to download from