Hacker News Reader

Comment by indymike | Hacker News Reader

I'm always amazed with how big tech companies get in terms of employee count. I once wandered the engineering area of the HQ (with an employee) of a FAANG (MAANG sounds... silly) and made a tick mark on my notebook when I saw someone actually coding - or at least having code on screen. There were over 1200 people on the development team and at the time most worked out of HQ. So... when I was done walking, I had 22 tick marks. Lots of email, lots of documents, and very little programming.

View on HN

Its almost like the practice of software engineering in a large organization involves (much) more than writing code?

Many of them are Software Bureocrats - their job is to keep organisational processes running with the goal of avoiding expensive mistakes. They are a lot like corporate lawyers. They don't create new stuff, but they save your organisation from descending into chaos.

Actual engineering still happens at the heart of the organisation, that heart is just shielded by the gigantic army of software bureaucrats.

More of them are akin to social workers enduring the morale never gets to expectant of rewards and the corporate sociopathy never disrupts too much

22 is less than 2% of the engineering effort. I agree that code part of the engineering effort, but if your business is software driven, you would think it is more than 2% of the effort. A lot of the adjacent roles to an engineer, i.e. documentation, devops, sre, all have code involved, too.

I get the sentiment here. But I can understand why someone would want to optimize towards writing more code than dealing with organizational politics. I guess the tradeoff would be to go to a smaller company.

Everywhere I've worked (including for MAANG companies), software engineering was best done by a combination of three things, listed in priority order:

A) thinking

B) programming

...

Z) meetings

The meetings contained something useful maybe 1/10th of the time, lasted maybe 4x longer than they should have, and involved maybe 20x more people than they should have. Those numbers skew higher the larger the organization.

Maybe that's a bit of an exaggeration, but not by much.

May we all be so fortunate to have tasks defined so specifically and selected so precisely that one can accomplish them only by thinking and programming, and find satisfaction in doing so.

Heh. I knew I’d get downvoted.

I’ve generally been on the most productive teams in my segment of the business. The biggest differentiator seems to be that my teams had devs that knew when to ask questions and what questions to ask. This is probably the most important skill a dev can learn.

The solution to ambiguous requirements is not meetings. At least, not in the usual sense. It’s knowing who and what to ask, and occasionally meeting with customers or knowledgeable stakeholders.

At a certain level of complexity, necessary to support everything a large company does at their scale, simply mindlessly churning out more code actually makes things worse not better. The big question is how to do something and not the actual work of doing it. Bringing down Facebook for a day due to a bad config is going to cost more money than a century of someone writing code.

A classic rant on the subject: Steve Yegge's "Code's Worst Enemy:"

https://steve-yegge.blogspot.com/2007/12/codes-worst-enemy.h...

>Facebook for a day due to a bad config is going to cost more money than a century of someone writing code.

I simply don't believe this is true. Its one of those things that has become "fact" through repetition. I don't think there is even a way it could be proven if you wanted to. I don't think amazon loses $X millions if the site is down for an hour. I think people just say oh its down and come back later and buy exactly what they were going to anyway.

Perhaps they have to pay out some SLA type stuff to advertisers. However I don't think there is a single outage of a major tech company that has lasted 24hrs in the last 20 years. Major tech company is of course debatable.

The reason they aren’t down more is because they have systems and processes to keep them up. It generally requires multiple failures for them to go down in any meaningful way.

That being said if your service went down for 24 hours because of a bad code change, it probably means and your service goes down a lot. Because whatever caused that 24hr outage probably points to a systemic lack of process and discipline within the organization.

Facebook was down for 6 hours last year during the day (in the western world) so that was very impactful most likely. And as wonderwonder said advertising is an eyeballs game and if no one is on the site then they're not seeing/clicking/attributing to ads. Any non-monopoly would have a decent impact from downtime as customers either spend their time on other sites or buy more time sensitive things on other sites.

edit: There's also a lot of revenue impacting things short of downtime especially given the automated and ML heavy paths for most of these sites. A few slightly broken features in a key model can cause a 1% drop in your revenue which for facebook is over a billion a year.

Facebook does not make money selling physical items though, it makes money on page views and clicks. If the site is down, they lose all of those ad views and clicks, they are gone because the time needed to view them and all of the other ads is gone.

I have not worked in the ad space so I'm certainly ignorant of the day to day mechanics. How can you prove that the people didn't just come back 6 hrs later and spend the same time clicking and viewing the same ads? Just in a different timeframe? Is the cost of an ad at 6PM really that much greater than at 10PM?

IME SLA's rarely pay out enough to really make it worth pursuing. However I have gotten a couple payouts from SLAs for companies i worked at in the past. I didn't think it was worth the time but I got paid either way.

I get they have profit models that "prove" they lost money but they all must be based on an assumption. The assumption that there was only this one chance to catch the attention of the customer and never again. To me this is a very big assumption that simply cannot be proven and is probably wrong.

>How can you prove that the people didn't just come back 6 hrs later and spend the same time clicking and viewing the same ads?

Why do you assume intelligent people who make decisions based on this haven't tracked spend by week, spend by month, regression models and the five hundred other ways to try and model this? We're talking about trillion dollar industries and you're assuming they don't do basic sanity checking.

I did start out saying I simply do not believe them. They don't have data from the time they were down, presumably. That is a big hole in the data for one instance.

>assuming they don't do basic sanity checking.

I'm sure they have a million ways to say its not my fault. I started writing a huge response about human nature and the incentive to massage stats but I think you will dismiss it. If a system is down its a great opportunity to get a blame hedge in case sales are not as good as you predicted. There is no incentive to take an honest look if a brief downtime really affected sales in a meaningful way. I doubt anyone even does look. They just say this it what it would have been in our models if the downtime did not happen.

dunning-kruger

Total Time Spent, Total Post Impressions, Total DAP are both easy to measure and relatively predictable week over week, so its fairly easy to see regressions. Outages definitely impact consumer apps, which makes sense if you stop to consider that a large chunk of time spent on platform comes during the downtime between other activities. Probably less so for commerce apps, especially ones that sell staple goods.

Think the extreme case - what if the site is down for a whole month, a whole day? Then you can understand.

Another way: you have nothing else to do at the time. so you might click the ads. But now the site is down, there is no ads for you to click.

I mean there was an article yesterday about people not buying/playing GTA5 because of a 5 minute load time.

Commercial web sites lose money when they are down. This is obvious to everyone.

For sure. My issue was that I think its way overestimated. They have every incentive from the sales side to pump up that number as much as possible.

Atlassian?

Most of these jobs hire software engineers, not programmers. The job is more than coding.

Software engineering is writing technical design docs. Gathering business requirements. Gathering technical requirements. Negotiating with stakeholders like InfoSec, legal and accounting. Discussing user research and new features with UX practitioners. Etc.

Mentally I always chuckle at MANGA (sleeve in Spanish, or the comic style),and it could also be MAGNA, but has possibly too good of a sound to it to be worth it.

Yes, but haven't you heard that productivity has gone up since we've moved to WFH? You'd probably have like 30-40 tick marks now if you toured all their homes.

Yeah i don’t think this is true though people love to say it is true

Productivity is a very subjective term

Are you making more dollars for less money? This is productivity, everything else doesn’t matter

I think a developer productivity is measured by the eng output, not the market pricing of the goods they sell.

Im pretty sure LOC/person went down a lot with the pandemic across all companies.

I would love to make more dollars for less money. I think.

Big MAMAA?

FAANG sounds silly too, we’re just used to it.

> (MAANG sounds... silly)

and mangy.

Reply on news.ycombinator.com