Why SQL server was never ported to Unix

Great article.

The decision to support anything new in the enterprise infrastructure space, given that you'll often be asked to support it for (literally) decades, is non-trivial even when you only have one platform to worry about. But especially when the second (or nth) platform in question has very different end-user expectations for usage conventions, documentation, scriptability, etc...it takes a very special partner opportunity (or end-customer so huge that it might as well be a partner opportunity) to make it worth pulling the trigger.

Databases in particular strike me as poor candidates, because the space is so hotly contested by organizations capable of research-class CS and deep optimization. So of all the things you might port to a new platform, MS SQL Server (a platform in itself) represents an enormous risk.

This is a great argument ... for open source.

Red Hat (the company I work for) sells quite narrow products. For example, we'll sell you and support PostgreSQL but only on Red Hat Enterprise Linux, or KVM but only on x86-64. Upstream, PostgreSQL runs on BSD, Windows and a zillion other platforms. KVM supports i386, S/390, PPC and a few others.

You can either be a customer, or you can port yourself / find another partner to make the code work on other platforms. This only works because it's open source software.

I don't doubt the author is reliably recalling what happened but I do doubt that Microsoft would have ever released an enterprise product for *nix. This would have eventually had to be decided at the Gates level and I don't think he would have ever undermined Windows with a tacit admission that it could not meet every customer need.

>> The first order engineering of making that happen, assuming you disable some Windows-specific functionality, was rather small (on the order of a few manweeks).

I have a hard time believing that. I used to write to the Windows internals API a lot (the kind of stuff you see on sysinternals), and I can't imagine the sql server team not doing the same. That stuff was not easily ported to Linux, at least in the late 90's / early 00's.

Slightly tangential - I didn't realize that this was Hal Berenson's blog. This is an individual who carried serious clout within MSFT. All his other posts are great reading too for an insider perspective on lots of things Microsoft.

Could someone pleae correct the title to read "Why Microsoft SQL Server was never ported to Unix" rather than the more general term SQL server (which Microsoft would no doubt be delighted to assume as its own...)

The most interesting part of this article to me was the description of the problem with sales. I think it provides a great insight into the thinking of large companies like Microsoft and why they're decisions are sometimes baffling to those of us not working at huge corporations. He mentions that a sales person going after $100k contracts would not make their quota and lose money for MS. I guess I shouldn't have been surprised but it seems pretty crazy.

It also kinda confirms what we already know - that every department of MS is geared towards pushing their whole software stack and to change would be like trying to redirect an iceberg.

Isn't the answer to that question trivial? Microsoft needed Windows monoculture to succeed and the way to do that is to completely and totally strangle any Windows customers so they could not move to something else without basically starting over.

Interesting exploration of the business issues, and, for Microsoft, I can understand this.

It does contrast rather strongly, though, with the experience of Oracle and Informix trialing their first Linux ports. "We just typed 'make'" was the quote out of Oracle at the time. While I'm reasonably confident that there has been a bit of platform-specific tuning since then, the point is that writing code for a standards-based platform (UNIX and POSIX) made porting to a new target pretty straightforward.

This is just sugar coating for the crap they're covering.

I'd reduce it all down to: we didn't want people to use SQL Server on *NIX because we wanted to sell them expensive OS licenses & licenses for other products.

>environment no one really knows nor cares what the underlying OS is. SQL Azure thus becomes Microsoft’s answer for those who don’t want to run an in-house Windows Server just so they can run SQL Server.

Nope, not really, I do care about the database and about the underlying OS. I don't like to use Windows for web hosting mainly due to its unpredictable HDD space usage (Windows folder growth vs number of installed updates over time).

Also, in Microsoft land, it's pretty difficult to be "up to date" with everything. You install hundreds of updates for the OS, the database and other components and there are some extra hotfixes you normally get when you have a support contract with them.

There's also the issue of the performance. You need more hardware for Windows than for other OSes. I had an "enterprise" CRM we will not name which had low performance on a Windows 2008 R1 Standard. I tried to upgrade it to R2 and the CRM stopped working completely. Of course, this wasn't done on the production box.

I could have sworn that way way way back when MS-SQLServer just started out it was a "fork" of Sybase Unix code??? I kinda remember that happening in like the 90's and then like usual Microsloth screwed it's partner (Then Sybase that did a lot of the work on the initial windows port)

Of course I maybe having some sort of strange drug induced flashback but kinda remember this?

RB

Wasn't that the database that used to regularly get it's indexes all mangled up?

I weep cold salty tears into my stale cheerios at the thought that any large Enterprise was ever seriously considering it as a viable solution.

Please stop calling MS SQL Server just "sql server". It's just as ridiculous as calling MS Internet Explorer just "internet".

Porting to unix? How about the other way: turning it OFF on windows?

sigh