Game_Ender

Can you explain why you don’t think stacked PRs should exist? What repo sizes, team sizes, and CI practices does your recommendation apply to?

I have seen them used in a few ways:

- To enable the creation of PRs that change just one thing. This makes them easier to review, revert or cherry pick.

- Allow you to continue to work while a more intensive CI or review process is under way.

With a good stacked change system, like in Phabricator or Gerrit, I have seen some of the most productive ever devs always using stacked changes.

This is Google's internal company wide issue tracker exposed to the public [0]. It was built requiring sign on because that is the default way a mature company builds tools so it can properly handle authz and authn.

My guess is nobody in that management chain cares enough to go through the effort of getting a mostly internal and collaborators system working without any auth.

0 - https://developers.google.com/issue-tracker

What percentage of writers, athletes, actors, or musicians are worth millions? Compare that to the number of STEM graduates.

It’s not that you cannot, it’t about the likelihood.

This is a pretty interesting project, a low setup distributed build cluster is cool. That said I find the arguments against Bazel pretty weak. Yes with a distcc model you can scale out the compilation stage of C++ project quickly, but what about the test execution, linking, ProtoBuf code generation, or caching? In either case, Bazel or Llama you have to make your build system aware of how these actions work in some way in order to distribute or cache them.

The difference is with Bazel you also gain a suite of solutions for large software projects like container creation, target visibility control, isolation, dependency fetching, and build graph analysis. Also with the larger community behind Bazel adopting it for a large project is getting easier over time.

I would simply stay away from Jenkins if you are getting started from scratch. I have used it for months, and the amount of effort you have to poor into it, and still get scaling and outage issues is not really acceptable.

In contrast when using Buildkite [0] you get essentially all the power and flexibility of Jenkins, but without the crushing technical debt, inherent flaws and complexity. Benefits I have seen of Buildkite over Jenkins:

- Never have to worry about scaling a Jenkins master again

- Build history lasts forever, so no need to setup a system to save logs

- Everything can be in your repo, so you are always confident of changes and testing is easy

- The UI is easy to understand and you can link directly to failed log lines

- Control your own workers, and easily setup autoscaling without any drama

- Flexible plugin and annotation system allows for extensibility

0 - https://buildkite.com/features

You can mitigate a bit of this by only ignoring blobs over a certain size like "--filter=blob:limit=256k" which should allow most ordinary text files through.

In the end it's the same as LFS though in that without a network examining old commits without a network is a bummer. No free lunch here besides something a bit more complex like git-annex.

The latest version of git has a very similar feature called “partial clones” to what the author describes for Mercurial. All the data is still in your history, no extra tools are needed, but you only fetch the blobs from the server for the commits you checkout. So just like LFS larger blobs not on master are effectively free, but you still grab all the blobs for your current commit.

You need server side support, which GitHub and GitLab have, and then a special clone command:

    git clone --filter=blob:none 

Some background about the feature is here: https://github.blog/2020-12-21-get-up-to-speed-with-partial-...

If Mozart had to spend more, or most, of his time gathering food and tending to his basic needs would he have had time to make music? What about the musicians who played it?

Wealth is the ability to do more with less effort. Some of that free labor goes to arts, science, or philosophy and advances humanity. Maybe too much goes toward building even more wealth or a fancier house/things but you would not have the former without the ability to get later.

You post PRs in a draft or WIP state and don’t request review until you are ready. This lets you leverage CI, but not take any time away from others who are doing review.

It was the Delta Clipper (DC-X) [0] and that was actually like the SpaceX, grasshopper. It stayed upright, translated hovered and came back to land.

Starship has flaps, a low terminal velocity belly flop maneuver, and then a rotation up to landing only a few hundred meters up. They definitely build on each other, but this is more impressive.

0 - https://en.m.wikipedia.org/wiki/McDonnell_Douglas_DC-X

Once the cutting has started the company is unlikely to have set the morale/team/event budget large enough to support go kart outings. That or the entire years budget would be gone at once.

Yes - https://www.washingtonpost.com/business/2019/07/24/doordash-...

If you are interested in this space and don't want to purchase a commercial product, check out Flagr [0] which you can deploy and run yourself. It has clients for Go, Python, Ruby and JavaScript. I have only used the basic flagging, but it also supports A/B testing as well.

No affiliation, just a happy customer.

0 - https://checkr.github.io/flagr

If you have the cycles porting the LLVM bazel build system to please and comparing performance and usability would be pretty sweet.

It seems like a great public use case/benchmark for a large common project to test with.

The world needs a Bazel package management system so the Bazel community can collaborate on projects like this. Buckaroo [0] is the closest I have seen but it’s Buck focused and they have not rolled out Bazel support.

I think ultimately it needs a company like npm that can upstream the needed changes to Bazel itself so that it supports these workflows better. With the right focus you could hit cargo levels of ease of use, but for large multi-language projects.

0 - https://github.com/LoopPerfect/buckaroo

You see the early versions of this effort every time you use the safari address bar. “Siri” will suggest a matching Wikipedia article or other direct link instead of you heading to google.

Leveraged properly all the OS level use from iOS could provide a lot of signal on which to build a search product.

If only Apple’s was able to do so for speech recognition.

I think Conan [1] has wider support in the C++ community. It definitely does not match the Cargo toolchain for integrated ease of use, but it does support CMake directly.

1 - https://conan.io/

Using less labor to grow food is what makes society productive enough for scientific and medical research. Those in turn address common public health issues and create more labor saving technologies allowing people to do things that they enjoy.

The camera was not in fact a direct upgrade, it was the same hardware with improved software post processing. That was the deciding factor for not getting an SE for me, and to wait for this cycle.

The real takeaway here: Honeycomb is a great tool for log and trace analysis, and by using Honeycomb's minimal data model [1] for traces you can expand tracing throughout much more of your stack. Now Honeycomb itself makes you pay by the event so Slack's internal buffer and collection system really lets them use it effectively for just the data they need. The low adoption of tracing is also due to the high cost of application level integration, which they are deferring by using the existing tracing systems service layer integration.

So if you come at this as a Honeycomb user Slack has not really re-invented any wheel so much built their own backend to funnel data to Honeycomb and use it's excellent trace viewing and light weight analytics UI [0]. In fact API and model which this article goes to great lengths to explain and justify is basically the Honeycomb tracing API and data model [1]. Which itself is an evolution of the system Facebook uses [2]. It is very cool that they have plugged the gap in Honeycomb which is the lack of historical analysis capability by also funneling all data into Presto.

I would love to see the system they use for keeping field/tag names under control. Honeycomb really strains under large numbers of field names.

0 - https://www.honeycomb.io/trace/

1 - https://docs.honeycomb.io/getting-data-in/tracing/send-trace...

2 - https://www.techrepublic.com/article/ex-facebook-engineers-l...

The drivers wanted in app tips. I think Lyft and Uber set rates assuming no tips, because usually people don’t tip and they get enough government heat already. They don’t want to end up like Door Dash and start using tips to explicitly cover base rates.

> dictate every aspect of the transaction

Uber let’s drivers set their own rates in CA [0]. What is also interesting is that before Lyft/Uber many taxi drivers were contractors renting their cars for an hourly rate. With rates set by the government, it’s really not much different than “rideshare”, just less flexible for them and the customer.

One issue is the type of market transport is and the variability in demand. You can do just a few Lyft trips, or just 1 and then do something else. CA employment law does not allow “micro shifts” like that, you have to pay the employees for a few hours. With no government restriction the market works itself out, with a large number of people who work a bit at peak supplementing those who are more full time. I think the legislature took the lazy way out, they could of set up rules for minimum compensation, benefits, And worker protection without the rest of the employment baggage.

0 - https://www.uber.com/blog/california/set-your-fares/

> I don't think I can tell someone else they don't deserve to earn minimum wage.

On the wage issue are you talking about the specifics of how prop 22 defines the earnings floor for drivers [0]? It definitely sets a standard even. It pays some for expenses ($0.30/mile) and the 120% minimum wage while on a ride covers a bit of time spent driving between rides. I could see though how if you have high expenses and don't get many rides you will not earn the minimum wage.

0 - https://ballotpedia.org/California_Proposition_22,_App-Based...

Uber drivers can set their own rates in some parts of CA [0]. It’s a pretty big change brought out by these lawsuits to make the drivers as independent as possible.

0 - https://www.npr.org/2020/01/28/800437791/due-to-new-californ...

You in effect pay people more to live in nicer areas. If I move to a place tons of other people move to it likely has some positives (better schools, more to do, closer to nice vacation opportunities), and higher costs.

Also if the high COL employee owns a property, much of that difference goes into equity in their home. Which they can they cash out and move to a low COL area.

It would really be interesting to learn more about their domain. Builds logs you have control over and should be able to keep clean and actionable. Service or batch job logs could present a more difficult problem though. It would be useful to try and spot a key message halfway through a failed 20 minute process that is the reason you hit an error at the end. Again though using logging levels and grep should usually be enough. You can also flag excessively noisy code and improve the signal to noise of the output it produces.

In specific with a build system like Bazel you might execute 1,000,000 actions on a build producing lots of internal output but you only see the errors, and at most you have a few hundred lines to look through. That is managed in a few key ways:

- Test output is completely hidden unless that specific test fails

- Build actions only produce output when they have an error, and it’s easy to keep it that way (because they stand out, and can be quickly fixed)

- Bazel does not tell you about every action that is run only ones that fail or take or a long time

Does anyone know why articles like this don't start all their graphs at 0? The information is persuasive enough without fudging things with poor axes. For example health expenditure as share of GDP graph starts at 8% and goes 18%, bounding the data very closely and magnifying the difference by several times.

A graph that start at zero

         Italy     USA
  0        *        *

How it was presented in the article:

  Italy     USA
    *        *

If you have a service or website you can (and should) do it dynamically with a service. I have seen Flagr [0] used to great effect here. As always you have to strip out flags when the features are permanent.

0 - https://checkr.github.io/flagr/#/

It’s clear from the article keeping the marketplace clean is not a company wide priority. Amazon famously has very decoupled teams so they have a team setup for this that filters incoming listings, and that is it. After all many of these banned items don’t really harm the user “just” society at large, so it’s not very “customer obsessed” to spend a big effort cross company to stamp them out. It would not surprise me if every flaw this article pointed out was known, and the isolated team simply does not have the people and politics power to fix it.

You can see this because the teams which manage virtually every other part of the site are not involved - the search recommendation was suggesting banned items - the buy together team does not use it’s system to find banned products related to removed items - the amazons choice team does not apply extra diligence to the things it flags. All things that a postmortem on now people find banned items would come up as action items. I bet they probably have, but it was low priory on those teams backlog.

It’s also clear they have not funded a big enough secret shopper team to just act like a user and try and buy banned products and then make reports to various teams across the site where they have gaps. If an outsider can sit down with your system and find bugs after a little bit of time, you don’t have a decent quality process in place.

How stateless and independent is each display?