To hear the New York Times’ John Markoff and Matt Richtel describe it in their largely fact-free story on the technical integration that Yahoo! and Microsoft will need to do if the merger goes through, you’d think that a Yahoo-Microsoft integration will amount to a cleaning of the Aegean stables.
The writers did take the time to interview someone who did a Unix-to-Microsoft port of a web site after it was purchased by Microsoft, but that port was done eight years ago. And this wasn’t Microsoft’s 1998 Hotmail acquisition (which some people consider to be the gold standard for Microsoft cocking up an acquisition of a *nix-based web property).
So the question is, from a technical integration perspective, could things have possibly changed in the past eight to ten years?
Well, of course they have. The one guy with direct knowledge that Markoff and Richtel interviewed (who now works for O’Reilly and should really know better) was probably migrating from Solaris to some version of Windows NT. One could imagine that LAMP as we know it today was not in the picture. And there are a broad spectrum of software engineering best practices that were not in place back then which go unmentioned in the article.
So ultimately, the piece leaves out several key facts that almost completely scuttle their thesis:
- Microsoft has worked with Zend to make PHP run well on Windows Server. In my February 1 post on the merger, I theorized that Microsoft did the Zend deal specifically to make it easier for them to digest companies (like Yahoo!) that extensively utilize PHP. The Markoff/Richtel piece does not mention the Microsoft/Zend deal at all. It does mention that PHP’s inventor works for Yahoo, but it’s a fact that Zend contributes far more to the language today and this has been the case for many years. At any rate, a reader of the Markoff piece could come away with the impression that PHP doesn’t run on windows at all, which is totally bogus.
- The hellacious Unix-to-Windows migrations of the late 1990s often hinged on the ability to get Oracle running on Windows, which was a big challenge then and remains a challenge today. But Yahoo uses very little Oracle in its customer-facing properties; the big database in use there is MySQL, which runs quite well on Windows today. I wouldn’t guess that Microsoft would migrate MySQL-on-FreeBSD to MySQL-on-Windows, but they could do it as an intermediate step if they wanted to get a merged Yahoo! running on Windows. But there is no mention of databases at all in the Markoff piece.
- FreeBSD, while prevalent at Yahoo!, is not actually the operating system that most companies in Silicon Valley (or anywhere) use. I’ve heard more than one Yahoo! engineer state that the company’s choice of that operating system is a hindrance for a number of reasons. Migrating to Anything But FreeBSD (whether it’s Linux, Windows Server, or Solaris) will have several key benefits — not the least of which being the fact that very few recent college grads are learning FreeBSD today. Another implication for this is that when Yahoo! acquires a company, they acquire that company’s operating system choice as well (all the company’s recent acquisitions, including Flickr, run on Linux, not FreeBSD).
- Both Yahoo! and Microsoft have made extensive investments in XML web services, both inside and outside the firewall, which should ease technical integration somewhat. You may recall that XML web services were considered cutting-edge and exotic in 1999. There is no mention of web services in the Markoff piece.
- The Windows Server that Microsoft sells today is not the piece o’ crap that Windows NT circa 1999 was. Period. The fact that Microsoft’s server technologies no longer suck is key, but it’s not mentioned in the Markoff piece.
- Probably the most ill-informed assertion is that Yahoo! is totally open source or that Microsoft is totally proprietary — or that either of these things would matter on a technical integration level even if they were completely true. But let’s imagine for a moment that someone needed to hack at the kernel of Windows Server to perform some Yahoo!/Microsoft integration task (which seems far-fetched, but let’s just imagine). Are we really to think that getting access to that code would be a deal-killing problem?