Slashdot Mirror
The Haps from LWCE: Samba Wins, RH w/XFS, BOF
We've been at LinuxWorld for the last couple days, and some interesting stuff has been going on:
The SAMBA folks won the $25,000 IDG/Linus Torvalds award, and SGI announced the availability of RH7-based distro using XFS [?] . In other news, our BOF went well with many questions about Slashcode - and the Perl Monks booth has been doing great in donations. Update: 02/01 05:18 PM by CT : The highlight for me so far was judging the "Coveted" Golden Penguin Awards w/ Don. Actually, I seriously did covet the award, beautiful hand blown glass penguin made me wish I was a contestant. We judged that Linus got the definition of BogoMIPS wrong. Fortunately his still won, but it was truly joyous seeing the surprise on his face.
Are the RLCBSDG (Red Latex Clad BSD Girls)there?
And where are the pix of said RLCBSDG?
---
This
In fact, it's just a modified installer for RedHat 7.0 that allows you to install it on XFS.
There are similar installers that make it possible to install RedHat, Debian and even Slackware on ReiserFS. More details on freshmeat.net
In short, it's not a separate Linux distribution from SGI.
Trade shows are pointless.
They're like big slow in-person websites without text. The only real good thing about them is that companies give you free trinkets, and from what I gathered, it wasn't all that much, although I did get a fuzzy little penguin guy from Penguin Computing.
I did give perlmonks.org some donation money (which consisted of a pile of money on a table), though. So that was kind of good, but otherwise, the whole thing was pretty boring. Even the people involved seemed quite bored.
Penguin awards ceremony and other webcasts from LW available here at Dr. Dobb's TechNetCast. Includes MP3s.
Shhhhhhh!
--
Liberty uber alles.
Um, bzzzt, wrong. SCO isn't pretty, or sexy, but they sell a LOT of licenses. They are a UNIX on Intel, and have a large business presence.
There servers have never been the most powerful (running on Intel Architecture, which until 2-3 years ago was seriously sucking compared to real Iron), but they have sold a lot of licenses.
The big Unix shops used to make a lot of money from licenses, but they were fewer licenses more $/license.
Not getting hype doesn't make you insignificant.
Hell, Solaris had limited presence in the Server room until the Internet boom, because their Hardware was considered substandard to IBM and HP stuff. However, they were a workstation player that gave their machines to engineering schools on the cheap, so people equated Sun with Unix.
Marketing gives the perception of marketshare, not the reality.
News.com has some video in RealVideo or Windows Media Player up now. They seem to be cutting over to a reporter live on the floor of the show, so I expect they'll be putting more up as they find their stories.
The slashdot stand looks kinda slick...
...j
SCO was hot 5-10 years ago. But since '95 or so they've been on the decline.
SCO sold licenses because there were all these various business tools that would only run on SCO.
About '96 or so most of these companies said "We're phasing out SCO in favor of X because it's easier/better/whatever".
Typically that X was Windows NT, sometimes it was Solaris.
I used to support an SCO ODT3 environment of about 8 workstations back in '93/'94.
But I agree with you in regards to Sun's popularity. Java also worked to their advantage, not necessarily as a language but as a marketing hype generator.
I'm running it exclusively on my home workstation at this point, FWIW.
---
Here's an interesting basis for comparison: I lived through two similar blackouts, one in a company that had mostly Sun servers, another that had mostly SGI servers. No warning in either case (the Sun installation had its UPSs taken out by a massive power surge just before the blackout; the SGI installation didn't even have UPSs).
In addition to the no-power time, the Sun-based installtion lost a couple days doing filesystem repair on all its system. The SGI-based installtion was back almost as soon as the power came back! An XFS disk is really a kind of database. As such, it's always in a consistent state. Changes to the filesystem are akin to database transactions. An incomplete transaction doesn't leave you with a broken filesystem -- you just lose the transaction. I know less about NTFS and ReiserFS, but I gather they have similar features. It's worth noting that XFS and NTFS didn't have defrag utilities for a long time. It's not that those are hard to write, it's just that fragmentation is supposed to be impossible on journalling FSs. Not in the real world, it seems.
__________________
I can't speak for performance (it feels fast, but I haven't used ext2 on the same hardware), but it feels pretty nice here. Install is currently a dump/restore affair, but I was building a new machine so just installed a base Debian system on a 150MB partition, copied over an XFS kernel, partitioned and then dumped stuff into the new root.
:)
I managed to nuke the primary superblock while attempting to install grub (I got my partition numbering confused), but thankfully there are recovery tools on SGI's website. It took about a minute to recover things from the secondary superblock. My only other problem was that 2.95 miscompiles the XFS code (you get long hangs during even moderately heavy file i/o), and I couldn't find any EGCS packages for Debian that didn't conflict with more recent ones. I ended up installing a gcc 2.97 snapshot and dropping back to 2.95 for the couple of files that gave me compiler errors. Still, it gives me a damn cool version string
Linux version 2.4.0-xfs (root@cavan) (gcc driver version 2.95.3 20010125 (prerelease) executing gcc version 2.97) #11 Wed Jan 31 11:41:20 GMT 2001
SGI are likely to promote Linux on their hardware. I don't see Irix being that long for the world in the lower-end workstation market based on a couple of conversations I've had with SGI people. If they port most of the useful features to Linux they get the advantage of a kernel that people will work on for free, letting them concentrate their efforts on the high-end multi-processor machines and media systems they make more money on.
This is great news!
Our research department made the decision to move from SGI Octanes at $30K a pop to cheap x86 boxes running Red Hat. Each SGI box has external storage between 8 GB and 36 GB.
Practical upshot: We can hang any disk attached to an SGI off a Red Hat 7 machine. Suddenly, our job of migrating from SGI to Linux just got a lot easier.
"I'm The Bounty Bear. I will find him anywhere. I'm searching."
Am I the only one who saw "Update: 02/01 05:18 PM by CT: " hours before 5:18 pm today? Taco got a time machine now?
Best Slashdot Co
As I understand it, XFS is usually used in big-iron servers. I was wondering what its relevance was to smaller servers and workstation systems. I have seen that it has a very large code base, and a ton of features, so what are its memory/disk-space/performance characteristics. Is it slower/faster/about-the-same as ReiserFS, does it make your kernel huge or what? It seems that some people on this board have used XFS, so any insights would be appreciated.
A deep unwavering belief is a sure sign you're missing something...
Which makes me wonder: why don't browsers send timezone information in their headers, so that the servers can do the Right Thing(TM) with regard to displaying times?
E.g. my browser should send a header like this:
TimeZone: America/Chicago
or
TimeZone: CDT/-600
Then the SlashCode could appropriately display the time that the story was updated as "11:18 AM". (Assuming that the "5:18" refers to GMT.
It seems like a fundamental need for web-based applications. Why hasn't this been done?
--
But didn't Linux first *define* BogoMIPS?
It's 10 PM. Do you know if you're un-American?
We are selling licences now? So this is how free software is going to make money: the software is free and free, but we sell the licences.
Eh, what would I DO with this licence I bought?Sig (appended to the end of comments I post, 54 chars)