Jim Gray - A talk with THE SQL Guru and Architect

This is well worth downloading and viewing more than once - full of ideas and genuine enthusiasm. Bring on Part 2!

Coming Monday.

Yeah, it was a big honor to be able to spend a few hours with Jim Gray and Gordon Bell and the other folks in the Bay Area Research Center.

So Jim has been researching grid computing for ten years, like before we even started calling it grid computing. That's pretty forward looking!

I like what he said about conventional OS design being a very narrow, limited field compared to what it could be. I've been saying to people for some time how depressing it is that all this effort goes into writing OSs that are all more or less the same as unix. I claim that virtually every mainstream OS since unix has been unduly influenced by it. People look at me like I'm mad, like Windows is nothing like unix they say. But the concepts are very similar. All OSs seem to have a similar concept of processes, of file streams; filesystems are nearly all hierarchical. Files are uniformly byte-adressable flat files from an OS perspective. Even if you go back to msdos you can see the influence of unix. Version 1 was more or less like CP/M in that the filesystem didn't have directories and files were accessed using what was called FCBs (File Control Blocks). Version 2 arrives and FCBs are still supported, but now you have a unix-style streams interface and a hierarchical filesystem.

Most of the credit for this goes down to the pre-eminence of not only the C language but more importantly the C library that defined the API. Later developments like the Berkley implementation of TCP/IP played a similar role in ensuring that almost all network programming is the same.

But yes, with a few notable exceptions like PICK and OS/400 it seems that the concept of an operating system for most people was set in stone by unix. Also the concept of the database seems to have been set in stone when Oracle version 1 shipped.

And that's just a comment on one of the first things Jim mentions. Then there's a heck of a lot of interesting things about science so much so that I almost forgot what I was going to comment on by the time I got to the end of the video!

scobleizer wrote:

Coming Monday.

Yeah, it was a big honor to be able to spend a few hours with Jim Gray and Gordon Bell and the other folks in the Bay Area Research Center.

Any chance you could tell me what else you have coming up on the same topic/genre?

V.interesting stuff...might even be able to base an entire session on it...very informative!

make me want more!! is that sad?

CSharpZealot wrote:

V.interesting stuff...might even be able to base an entire session on it...very informative!

make me want more!! is that sad?

Yes it is...... but you're among other sad people (very happy sad people, mind you)

I could spend days, leaving customers without support, watching this stuff :>

didn't have one until now.. but channel9 makes me buy a dvd burner. i'd have to say.. this interview is definitely among the best on the site. thanks!
i think i'll have to watch it once more to really get everything out of it.

great to hear such a seasoned mind; what's with the cut-outs in video though (especially when talking about the flaws of the MSDN site)? are these videos at all audited by microsoft PR?

i've noticed some snips to the video reel in a couple of these videos, some of them i can tell are just for asthetical and time-conservation purposes, but some of them seem like audits...?

Of course at almost 1 hr long...They might have had to cut somethings out. I am also sure that the content gets filtered in some way. Geeze it would have to and not just because Micosoft is who they are, any company in a simillar situation would be foolish NOT to at least examine such content.

A fascinating listen though. I definately downloaded it just to watch it again...

One of the top 3 videos I've seen on C9. I've also just learned that NASA's WorldWind uses data from Terra Service. Thanks, Jim, for providing a real entertaining & informative afternoon.

Maybe we should encourage Jim to think seriously about targeting his creativity and intellect on the future of operating systems in the connected world.

100% Charles!  If you think about it, WinFS is the start of the DB being directly integrated in the "OS".  If you think about the new features in SQL 2005, you could also start thinking on how to apply some of those features into the OS.  For example.  Say everything (not all) is now a job put into a queue (i.e. mainframe style).  Even common things like file copy and print jobs.  Each Windows OS will have a Service Brocker (like SQL 2005).  So file copy jobs and print jobs will be sent to the OS's Service Brocker Queue first instead handled directly.  Then it can better handle things like priority and resource usage with rules on the Queue and have commit and rollback symantics.  Think about things like tranfering files, xml documents, print jobs, etc between partners.  Just queue up a bunch of stuff and pool a reply queue or get some event notification.  Kinda like msmq, but directly integrated into the OS, reliable, available, and abstracted with the OS apis and GUIs.  Maybe email of the future would be xml documents handled by Service Broker queues between OSes with no SMTP.  Inbox would then just pull from respective queue, same with outbox.  Lots of cool things to think about.  

"Think about things like tranfering files, xml documents, print jobs, etc between partners.  Just queue up a bunch of stuff and pool a reply queue or get some event notification.  Kinda like msmq, but directly integrated into the OS, reliable, available, and abstracted with the OS apis and GUIs.  Maybe email of the future would be xml documents handled by Service Broker queues between OSes with no SMTP.  Inbox would then just pull from respective queue, same with outbox.  Lots of cool things to think about."

Now that I think about it more, we just solved all or most the worlds distributed issues.  No more network issues such as web services, sockets, or WSE for developers to deal with.  All are gone at this level.  Just need a DNS name, a Queue Endpoint Name, and your XML document (or other).  Send it to the queue and wait for reply via various means.  You get the reply or some error message.  Works with files, documents, binary, etc.  You query the SO contracts the same way (e.g. wsdl). The OS service broker handles the rest under the covers (i.e. federated security, network, protocols, etc.)  Boy that was easy.  Next!

For those keeping track at home, Jim is the godfather of TP as we know it and an ACM Turing award winner. Jim is also quite an influence to many of us on the WinFS team (and the SQL group at large).