The funny thing about this list is that a few of the items, while intended as satire, are actually spot-on. Especially item #1:
"But the rate of source code commits continues to grow proportionally with the number of engineers. This is in clear violation of the law that Fred Brooks established nearly 40 years ago in The Mythical Man Month."
Whoop de do.
This might, just maybe, because no particular behavior of FB really matters all that much. You're not building a space shuttle launch program or even a spreadsheet that people rely on to give correct answers.
I'm sure that a business model that can allow 80 godzillion developers to fling spaghetti at a wall in parallel with a view to seeing 'what sticks' is quite a bit less subject to Mythical Man Month type scaling problems.
And is there anyone so naive that they don't understand the agenda behind "3 free meals"?
It's likely that the code is not coupled. Fred Brooks studied OS/360, which was essentially one big component. Facebook is, presumably, many separate components that don't depend on each other in any way, and so the "teams" are much smaller than "all of Facebook".
If you watch velocity on individual components, you'll probably see sub-linear scaling as Brooks predicts.
The messaging back-end is probably completely decoupled from the stuff running other components of the site. I would imagine everything interfaces with each other, and as long as you don't break the interface, things could be changed pretty orthogonally.
Indeed. They have quite a toolchain and a wide variety of libraries. While a change in the javascript framework or PHP compiler affects everything, it isn't dependent on everything.
"Facebook is, presumably, many separate components that don't depend on each other in any way"
Based upon the public statements about their build system and their revision control system and pain when trying to put it in Git I would be confident in saying that is false and that the source code is one big intertwined ball and no separate components.
Edit: Also based upon the way they treat public API which is always better than internal API even if they had internal separate components one could only imagine the horrors of api design (or lack of) you might find.
Question is how well peer-reviewed the code is before deployment. Facebook is handling a lot of personal data about their users. Security vulnerabilities may not harm Facebook as much as their users.
"But the rate of source code commits continues to grow proportionally with the number of engineers. This is in clear violation of the law that Fred Brooks established nearly 40 years ago in The Mythical Man Month."
Whoop de do.
This might, just maybe, because no particular behavior of FB really matters all that much. You're not building a space shuttle launch program or even a spreadsheet that people rely on to give correct answers.
I'm sure that a business model that can allow 80 godzillion developers to fling spaghetti at a wall in parallel with a view to seeing 'what sticks' is quite a bit less subject to Mythical Man Month type scaling problems.
And is there anyone so naive that they don't understand the agenda behind "3 free meals"?