Three comments:
1) The Register is to 'news' as the New York Post is to the New York Times, which is to say its amusing at times and sometimes a good lead into something that is actually important, but rarely the definitive source for any story.
2) The entrepreneurial take away is that Web 2.0 has been racing ahead of hardware, its been able to do that because computers got better faster and iterating has proven more valuable that polishing.
There are opportunities to be found by taking a good long look at where the technologies used can benefit from abstraction/compilation love. Remember that in 2005 putting together a 8,000 'node' cluster was really on the fringe, today that is 500 boxes with a dual Core i7 motherboard in them.
3) In case you haven't noticed this reads like a Facebook Fanboi insulting a Google Fanboi. Apparently it was modestly successful in this regard if you measure success by vitriol in the comments on the article.
I think it's the output. Also, the timing is relevant as the project documentation keeps talking about distributed compiling etc. I think that's how the sentence was meant: we can compile 1GB in 15 minutes (speed).
Is this actually inaccurate though? Google has released tons of open source software, but which of them are important pieces of its back-end infrastructure? Hadoop and all its sub-projects are inspired by Google papers but aren't based on Google code, as far as I know. The projects that come to mind are Sawzall (of dubious usefulness to people outside Google), gold linker, possibly Go language (but I don't know to what extent that's used within Google), and what else? Probably a lot of contributions to open-source that it uses, like the Linux kernel, but I'm after whole projects created by Google. Python doesn't count because it was already open source before they acquired GvR.
Compare to Facebook, which has open-sourced Cassandra, Hive, HipHop, Thrift, Tornado, and others which I may be forgetting. Seems to me that Facebook is way ahead.
Edit: just remembered a Protocol Buffers, that's a good one to put under Google's column.
Right. None of which you could reasonably describe as "the most important pieces of its back-end infrastructure," unlike HipHop, Thrift, HBase, Cassandra, memcache, etc. at Facebook. Google has contributed a ton of wonderful products to open source; however they will never contribute BigTable, MapReduce, GFS, Chubby, GoogleBot, their indexing pipeline, the actual reverse index query execution engines, the search engine aggregators, Spanner, etc.
If you think about it, this is all perfectly sensible. Google sees back-end technology as a key competitive differentiator. Facebook needs to build back-end technology, too, but is not as vulnerable to a technologically equal competitor; if we were, Google would have made good on their promise to destroy us by now.
Disclaimer: I'm a Facebook engineer who contributes to the HipHop effort.
i guess thats all true. i think the release of the go language will prove to be a huge milestone for google both in terms of open source releases and in terms of projects by googlers.
So they dropped some rarely used functionality from PHP,... like serialization.
Excuse me, "rarely used"?? Drupal strongly depends on serialization, to store data structures in the database (or, as they call it, "the cache").
You cannot effectively run Drupal without serialization. So "The likes of Drupal, MediaWiki, and WordPress are now using HipHop." sounds like it cannot be true.
Except APC doesn't actually serialize anything except for objects. Arrays and scalar types are directly memcpy'ed in and out of the APC user cache. But this author is extremely confused. I think he may have confused APC with Memcache here. They have done some work on the PHP Memcache extension to remove the need to serialize making it behave more like APC.
I think Facebook is worried more about scale, hence HipHop. Drupal doesn't scale well at all and shouldn't be used as an example of what's best practice. Otherwise we'd see tons of products having function chains longer than a tapeworm.
2) The entrepreneurial take away is that Web 2.0 has been racing ahead of hardware, its been able to do that because computers got better faster and iterating has proven more valuable that polishing.
There are opportunities to be found by taking a good long look at where the technologies used can benefit from abstraction/compilation love. Remember that in 2005 putting together a 8,000 'node' cluster was really on the fringe, today that is 500 boxes with a dual Core i7 motherboard in them.
3) In case you haven't noticed this reads like a Facebook Fanboi insulting a Google Fanboi. Apparently it was modestly successful in this regard if you measure success by vitriol in the comments on the article.