FAQ

Neat site: Advogato

Chip Salzenberg
Apr 4, 2000 at 8:18 pm
I'm signed up on advogato.com, and I like it. I'm thinking that its
diary feature would be a good place to keep y'all in touch with what's
happening in development between snapshots....
--
Chip Salzenberg - a.k.a. - <chip@valinux.com>
"I wanted to play hopscotch with the impenetrable mystery of existence,
but he stepped in a wormhole and had to go in early." // MST3K
reply

Search Discussions

44 responses

  • John Tobey at Apr 4, 2000 at 8:50 pm

    Date: Tue, 4 Apr 2000 13:19:07 -0700
    From: Chip Salzenberg <chip@valinux.com>

    I'm signed up on advogato.com, and I like it. I'm thinking that its
    diary feature would be a good place to keep y'all in touch with what's
    happening in development between snapshots....
    Well it's better than nothing, although CVS over SSH is nicer IMHO if
    it suits your development style. (sourceforge supports it, btw)

    -John
    --
    Chip Salzenberg - a.k.a. - <chip@valinux.com>
    "I wanted to play hopscotch with the impenetrable mystery of existence,
    but he stepped in a wormhole and had to go in early." // MST3K
    --
    John Tobey, late nite hacker <jtobey@john-edwin-tobey.org>
    \\\ ///
    ]]] With enough bugs, all eyes are shallow. [[[
    /// \\\
  • Chip Salzenberg at Apr 4, 2000 at 9:19 pm

    According to John Tobey:
    I'm signed up on advogato.com, and I like it. I'm thinking that its
    diary feature would be a good place to keep y'all in touch with what's
    happening in development between snapshots....
    Well it's better than nothing, although CVS over SSH is nicer IMHO if
    it suits your development style.
    Ouch! That stang. :-)

    Well, actually, I'm probably pretty close to cvs-ability. The issue
    is directory layout and files being created/deleted. But I can get
    the SourceForge folks to delete my CVS tree if I have to start over, I
    guess.
    --
    Chip Salzenberg - a.k.a. - <chip@valinux.com>
    "I wanted to play hopscotch with the impenetrable mystery of existence,
    but he stepped in a wormhole and had to go in early." // MST3K
  • John Tobey at Apr 4, 2000 at 9:26 pm

    Well it's better than nothing, although CVS over SSH is nicer IMHO if
    it suits your development style.
    Ouch! That stang. :-)
    Sorry! :-)
    Well, actually, I'm probably pretty close to cvs-ability.
    I'm willing to help.
    The issue
    is directory layout and files being created/deleted.
    I bet a "cvs mv" that I wouldn't be ashamed of could be implemented in
    50 lines of Perl. (is that what you meant?)
    But I can get
    the SourceForge folks to delete my CVS tree if I have to start over, I
    guess.
    Ah the perks!
    --
    Chip Salzenberg - a.k.a. - <chip@valinux.com>
    "I wanted to play hopscotch with the impenetrable mystery of existence,
    but he stepped in a wormhole and had to go in early." // MST3K
  • Chip Salzenberg at Apr 4, 2000 at 10:06 pm

    According to John Tobey:
    I'm willing to help. Thanks!
    I bet a "cvs mv" that I wouldn't be ashamed of could be implemented
    in 50 lines of Perl. (is that what you meant?)
    I think so. I'd like to see it, actually. And if it copes properly
    with the target directory not yet existing, bonus points.
    --
    Chip Salzenberg - a.k.a. - <chip@valinux.com>
    "I wanted to play hopscotch with the impenetrable mystery of existence,
    but he stepped in a wormhole and had to go in early." // MST3K
  • Russ Allbery at Apr 4, 2000 at 10:20 pm

    Chip Salzenberg writes:
    According to John Tobey:
    I bet a "cvs mv" that I wouldn't be ashamed of could be implemented
    in 50 lines of Perl. (is that what you meant?)
    I think so. I'd like to see it, actually. And if it copes properly
    with the target directory not yet existing, bonus points.
    This is a topic of regular and very heated discussion on info-cvs, just so
    that you're warned. It's rather difficult to come up with good semantics
    for file moves that don't have various drawbacks.

    --
    Russ Allbery (rra@stanford.edu) <http://www.eyrie.org/~eagle/>
  • Bennett Todd at Apr 4, 2000 at 10:32 pm

    2000-04-04-18:19:53 Russ Allbery:
    Chip Salzenberg:
    Iif it copes properly with the target directory not yet
    existing, bonus points.
    This is a topic of regular and very heated discussion on info-cvs,
    just so that you're warned. It's rather difficult to come up with
    good semantics for file moves that don't have various drawbacks.
    Understandable, since the CVS central trick, the heart around which
    all the other frippery is built, revolves around identifying
    differences in a useful and meaningful fashion. Like rsync in a way,
    another program that doesn't currently do the Smart Thing with file
    moves, and that might profit, in some applications, if it could.

    Pretty much any really worthwhile file-move support, intended to be
    an improvement over delete/create, could force CVS to abandon it's
    current representation using stock RCS files, or at least leave that
    representation incomplete, which would be a shame.

    -Bennett
  • Chip Salzenberg at Apr 5, 2000 at 4:04 am

    According to Bennett Todd:
    Pretty much any really worthwhile file-move support, intended to be
    an improvement over delete/create, could force CVS to abandon it's
    current representation using stock RCS files [...]
    Well, it's pretty obvious when you consider why Windows has to reboot
    to install DLLs.

    Windows doesn't have inodes. So Windows can't cope with a separation
    between name and identity. Unix, on the other hand, is entirely happy
    to have N different files open simultaneously, all of which may have
    shared a common name at some point.

    CVS has an analogous dependency on filename indicating identity. More
    advanced systems like ClearCase track file identity separately from
    filename, allowing easy and natural support for renaming files.

    In other words:

    CVS is to ClearCase as Windows is to Unix.

    <donning asbestos>
    --
    Chip Salzenberg - a.k.a. - <chip@valinux.com>
    "I wanted to play hopscotch with the impenetrable mystery of existence,
    but he stepped in a wormhole and had to go in early." // MST3K
  • Russ Allbery at Apr 5, 2000 at 4:44 am

    Chip Salzenberg writes:

    CVS has an analogous dependency on filename indicating identity. More
    advanced systems like ClearCase track file identity separately from
    filename, allowing easy and natural support for renaming files.
    Yes, but there's a reason for this. This is because CVS is nothing more
    or less than a system to help manage parallel RCS files, and RCS is a
    per-file revision control system. CVS has honestly gone about as far as
    it can and still maintain that position, and I'm one of the people who
    thinks that it's just about perfect for a lot of what I do. Not the least
    because I have tons of old RCS revision histories that are absolutely
    trivial to import into CVS because of it. :)

    Renaming files is a significant wart; the ease, standardization, and
    simple structure of the repository more than makes up for it IMO.

    --
    Russ Allbery (rra@stanford.edu) <http://www.eyrie.org/~eagle/>
  • Chip Salzenberg at Apr 5, 2000 at 7:54 am

    According to Russ Allbery:
    [...] Not the least because I have tons of old RCS revision
    histories that are absolutely trivial to import into CVS because of
    it. :)
    Well, now that you mention it, Deliver's RCS history would be easy to
    put on SourceForge. Hm. Well, I guess it's not evil after all.
    --
    Chip Salzenberg - a.k.a. - <chip@valinux.com>
    "I wanted to play hopscotch with the impenetrable mystery of existence,
    but he stepped in a wormhole and had to go in early." // MST3K
  • Joshua N Pritikin at Apr 5, 2000 at 11:51 am

    On Tue, Apr 04, 2000 at 09:03:51PM -0700, chip@valinux.com wrote:
    CVS has an analogous dependency on filename indicating identity. More
    advanced systems like ClearCase track file identity separately from
    filename, allowing easy and natural support for renaming files.

    In other words:

    CVS is to ClearCase as Windows is to Unix.

    <donning asbestos>
    Or perforce... ;-)

    --
    "Never ascribe to malice that which can be explained by stupidity."
    via, but not speaking for Deutsche Bank
  • John Porter at Apr 5, 2000 at 4:01 pm

    Chip Salzenberg wrote:

    CVS is to ClearCase as Windows is to Unix.
    I agree 200%; ClearCase is a brilliant product. More people
    should know aobut it. (And it should be less expensive.)

    OTOH, saying CVS is in any way like Windows makes me want
    to avoid it...

    --
    John Porter
  • Bennett Todd at Apr 5, 2000 at 4:21 pm

    2000-04-05-12:01:13 John Porter:
    OTOH, saying CVS is in any way like Windows makes me want
    to avoid it...
    I think that was the intent. Unfortunate, too; I think CVS is a
    superb tool. It's not perfect, and it doesn't do everything there is
    --- or try to. What's more, it's not especially user-friendly.

    But I really like its balance of complexity -vs- functionality; I
    like the way it exposes as much as you want of its underpinnings. By
    sticking strictly to a directory heirarchy of RCS files, it may end
    up with limitations that turn icko when you want to rename things
    (although create and remove isn't an awful way of handling it). But
    it retains lovely compatibility with other tools.

    I find CVS a _really_ robust foundation onto which to layer other
    tasks. I check every backup of my pocket computer (was an HP LX, is
    now a Palm) into a CVS tree right after I take it, automatically
    from the same script. I park a CVS tree in the scripted data flow
    between staging and production webservers.

    I think the implication that CVS is like Windows is pretty much pure
    flame-bait.

    In fact, if you were gonna claim that a version control system is
    like Windows, wouldn't it make more sense to look for a proprietary
    system, distributed in binary form only, whose innards are
    undocumented and inaccessible?

    -Bennett
  • Michael G Schwern at Apr 5, 2000 at 5:04 pm
    Just to throw my $0.02 of gasoline on the fire... there are two things
    which keep me on CVS (well, maybe three).

    1) Its free.
    ClearCase is great and all, but its bloody expensive. The offshoot
    of this is that almost everyone has (or can get) a CVS client. Not
    everyone wants to (or can) buy a ClearCase client. This is, of course
    extremely important in an Open Source project.

    2) Its text.
    Now matter what horrific things you do to it, you can
    almost always recover a CVS repository. No mysterious binary file
    formats. No registry to corrupt. If worse comes to worse you can
    just hack around in the cvs root.

    3) It links into emacs' vc-mode. :)
    Although I'm not sure if the big commercial version control systems
    do or not. Haven't looked. I think ClearCase does.

    That being said, my garbage list has gotten rather long in the one
    year that I've used it. Lack of support for symlinks, hairy method of
    moving files, almost impossible to remove or move a directory (without
    doing it in the root itself), silly defaults for commands, massively
    overloaded interface (HOW many options does "cvs update" have?)

    As kind of an offshoot, I've done a little bit with SourceForge and I
    found the greatest lacking to be the inability to get at the root
    directly. I've become so accustomed to having to do stuff manually
    that having only cvs/ssh access to my repository leaves me hamstrung.

    However, I do applaud CVS's staunch decision to be a version control
    system and ONLY a version control system. There's too many
    version/build/config/manage/test/distribute bloated all-in-wonders out
    there already. Its nice to see a project which hasn't succumbed to
    the urge to do everything.


    PS Has anyone experience with Bitmover? Larry ummm... Rosler is it?
    was bubbling over about it at the Linux Expo '98 but I never got a
    chance to really look at it.


    PPS As an amusing aside, I remember talking to one of the commercial
    version control guys (I don't want to name names because I don't
    remember exactly who it was) in early '98. Their salesman came down
    and was showing us all the wonderful things their system could
    do... on a Windows box. Apparently there was no Unix client, just a
    Solaris server and a Windows client. When pressed he said that they
    didn't want to get involved with Linux or BSD because there's no tech
    support for them and they're too unstable anyway.

    We've come a long way, baby.


    --

    Michael G Schwern http://www.pobox.com/~schwern/ schwern@pobox.com
    Right Wing Enema:
    Flush Immorality Now!
    God gums gay gorgings.
    -- boojum
  • Chip Salzenberg at Apr 5, 2000 at 6:09 pm

    According to Michael G Schwern:
    ClearCase is great and all, but its bloody expensive.
    Yeah. And if only there were a free version ... Shoot, maybe that's
    what I'll do when I'm done with Topaz.
    PS Has anyone experience with Bitmover?
    That would be BitKeeper ... and no, I haven't used it yet.
    --
    Chip Salzenberg - a.k.a. - <chip@valinux.com>
    "I wanted to play hopscotch with the impenetrable mystery of existence,
    but he stepped in a wormhole and had to go in early." // MST3K
  • Gerd Knops at Apr 5, 2000 at 6:22 pm

    Chip Salzenberg wrote:
    According to Michael G Schwern:
    ClearCase is great and all, but its bloody expensive.
    Yeah. And if only there were a free version ... Shoot, maybe that's
    what I'll do when I'm done with Topaz.
    PS Has anyone experience with Bitmover?
    That would be BitKeeper ... and no, I haven't used it yet.
    What's wrong with Perforce? Not open source, but they provide free
    FreeBSD/Linux licenses for developers of free software.

    I never could get the hang of CVS, it's just to darn complicated and
    to easy to do the wrong thing. Perforce is much easier to use, and
    has been working reliable for me and many others for a long time.

    Gerd
  • Nick Ing-Simmons at Apr 5, 2000 at 8:10 pm

    Gerd Knops writes:
    What's wrong with Perforce? Not open source, but they provide free
    FreeBSD/Linux licenses for developers of free software.

    I never could get the hang of CVS, it's just to darn complicated and
    to easy to do the wrong thing. Perforce is much easier to use, and
    has been working reliable for me and many others for a long time.
    Seconded. It works really well over ssh - even with a dubious 28K modem
    link it was adequate to work on perl5.005 when depot was in Oxford
    and I was in Dallas. Still fine with depot in Vancover and me in UK.
    Gerd
    --
    Nick Ing-Simmons
  • Chip Salzenberg at Apr 5, 2000 at 8:46 pm

    According to Gerd Knops:
    What's wrong with Perforce? Not open source, but they provide free
    FreeBSD/Linux licenses for developers of free software.
    Well, hm, if we could get a site license for SourceForge ... Hm,
    that's an interesting idea!
    --
    Chip Salzenberg - a.k.a. - <chip@valinux.com>
    "I wanted to play hopscotch with the impenetrable mystery of existence,
    but he stepped in a wormhole and had to go in early." // MST3K
  • Russ Allbery at Apr 5, 2000 at 9:07 pm

    Chip Salzenberg writes:
    According to Gerd Knops:
    What's wrong with Perforce? Not open source, but they provide free
    FreeBSD/Linux licenses for developers of free software.
    Well, hm, if we could get a site license for SourceForge ... Hm,
    that's an interesting idea!
    I thought one of the points of SourceForge was that it was run on 100%
    free software. That's something I respect a lot about VA Linux.

    --
    Russ Allbery (rra@stanford.edu) <http://www.eyrie.org/~eagle/>
  • Chip Salzenberg at Apr 5, 2000 at 9:28 pm

    According to Russ Allbery:
    I thought one of the points of SourceForge was that it was run on 100%
    free software. That's something I respect a lot about VA Linux.
    Ah, you're right. That's probably why they aren't supporting
    BitKeeper. Oh, well.
    --
    Chip Salzenberg - a.k.a. - <chip@valinux.com>
    "I wanted to play hopscotch with the impenetrable mystery of existence,
    but he stepped in a wormhole and had to go in early." // MST3K
  • Joshua N Pritikin at Apr 6, 2000 at 12:03 pm

    On Wed, Apr 05, 2000 at 01:46:24PM -0700, chip@valinux.com wrote:
    According to Gerd Knops:
    What's wrong with Perforce? Not open source, but they provide free
    FreeBSD/Linux licenses for developers of free software.
    Well, hm, if we could get a site license for SourceForge ... Hm,
    that's an interesting idea!
    I actually begged the perforce people to talk to the people at
    SourceForge. They seemed to have the impression that SourceForge wasn't
    willing to work with non-open source software. But why can't the
    developers decide? Chip, can't you do something?

    --
    "May the best description of competition prevail."
    via, but not speaking for Deutsche Bank
  • Russ Allbery at Apr 6, 2000 at 9:39 pm

    Joshua N Pritikin writes:

    I actually begged the perforce people to talk to the people at
    SourceForge. They seemed to have the impression that SourceForge wasn't
    willing to work with non-open source software. But why can't the
    developers decide?
    As I understand it, it would violate one of the core points behind
    SourceForge. Remember, VA Linux is a free software company; that's the
    market position that they're taking, and they're trying to convince the
    conventional world that it's a viable market position to take. One of the
    things that they get in return for donating the impressive resources of
    SourceForge to the community is publicity for their approach.

    I really agree with a position of using only free software; I think it's a
    good standard to try to keep. It proves something to the rest of the
    world, which is in some ways more important than the minor annoyances of
    using one VC system over another. Maybe ten or twenty years down the
    road, it won't be necessary to prove this point any more because people
    will understand that free software can be good software and it won't hurt
    the message to just use whatever's the most convenient for the job, but I
    don't think we're there yet.

    Unfortunately.

    --
    Russ Allbery (rra@stanford.edu) <http://www.eyrie.org/~eagle/>
  • Chip Salzenberg at Apr 7, 2000 at 12:35 am

    According to Russ Allbery:
    I really agree with a position of using only free software; I think
    it's a good standard to try to keep.
    I have to agree. It may not be reasonable to enforce an all-free
    policy on any arbitrary person or group, YET. But one of the open
    source community's strong points is -- no surprise -- software for
    software development. So in that area, yes, all-free is OK.

    Granted that limits our choices, but hey, it's making me think that we
    really need a good open source feature-a-like of ClearCase. And if we
    end up with that, or something else good, as a result of SourceForge's
    stand, well, maybe that was the whole point.
    --
    Chip Salzenberg - a.k.a. - <chip@valinux.com>
    "I wanted to play hopscotch with the impenetrable mystery of existence,
    but he stepped in a wormhole and had to go in early." // MST3K
  • Adam Turoff at Apr 5, 2000 at 7:19 pm

    Chip Salzenberg wrote:

    According to Michael G Schwern:
    ClearCase is great and all, but its bloody expensive.
    Yeah. And if only there were a free version ... Shoot, maybe that's
    what I'll do when I'm done with Topaz.
    I honestly don't see the point in that.

    CVS is a good default VCS. Seeing how many developers have gone so very
    long without any real version control, CVS is a godsend, even with all
    of its warts, flaws and limitations. Calling it garbage, comparing it
    to Windows and such is really counterproductive. Sure, it tops out when
    you try and push the envelope. Sure it has problems scaling. But it
    serves as a valueable "starter" VCS, a lingua franca that anyone can use
    (and probably has used if they've moved to BitKeeper, Perforce or ClearCase),
    and a quite reasonable baseline for VCS today.

    Now, if you want a free ClearCase, start lobbying the vendor.

    If you want a better CVS than CVS, there are already products out
    there that are free and possibly Open Source.

    Making Topaz++ a better Topaz would be a more worthwhile goal.
    PS Has anyone experience with Bitmover?
    That would be BitKeeper ... and no, I haven't used it yet.
    I'm pretty sure that BitKeeper is free for Open Source projects. I
    imagine BitMover would be quite happy to drop BitKeeper on sourceforge
    for anyone who wants it, if only because of the marketing value.

    http://www.bitkeeper.com
    http://www.bitmover.com

    -- Adam
  • John Porter at Apr 5, 2000 at 6:42 pm

    Michael G Schwern wrote:
    Just to throw my $0.02 of gasoline on the fire... there are two things
    which keep me on CVS (well, maybe three).

    1) Its free.
    2) Its text.

    Lack of support for symlinks, hairy method of
    moving files, almost impossible to remove or move a directory (without
    doing it in the root itself),

    However, I do applaud CVS's staunch decision to be a version control
    system and ONLY a version control system. There's too many
    version/build/config/manage/test/distribute bloated all-in-wonders out
    there already. Its nice to see a project which hasn't succumbed to
    the urge to do everything.
    All points on which CVS meaningfully differs from ClearCase.
    I wouldn't even mention it, except that mv'ing in CVS seems to be
    such a big deal: mv'ing a file in ClearCase is mindblowingly easy;
    it does the rmelem and mkelem automatically (of course, the source
    and destination dirs have to be checked out).

    The most brilliant thing about ClearCase is the mvfs -- multi-versioning
    file system. It's so powerful and so transparent, I'm always tempted
    to keep *everything* -- down to my mboxes and .kshrcs -- in it.
    (But I don't...)

    --
    John Porter
  • John Tobey at Apr 5, 2000 at 7:25 pm

    John Porter wrote:
    I wouldn't even mention it, except that mv'ing in CVS seems to be
    such a big deal: mv'ing a file in ClearCase is mindblowingly easy;
    it does the rmelem and mkelem automatically (of course, the source
    and destination dirs have to be checked out).
    OK, here's a band-aid:

    http://john-edwin-tobey.org/Perl/cvsmv

    Here's the documentation:

    Usage:
    cvsmv FILE NEW - do this after mv FILE NEW
    cvsmv DIR NEW - do *NOT* first mv DIR NEW
    cvsmv ... DIR - do this after moving files, but not dirs, to DIR

    You move the files, I move the directories, I move stuff in CVS.
    I commit every move so I can add appropriate log messages.

    Sorry, it's not in CVS. ;-)

    --
    John Tobey, late nite hacker <jtobey@john-edwin-tobey.org>
    \\\ ///
    ]]] With enough bugs, all eyes are shallow. [[[
    /// \\\
  • Michael G Schwern at Apr 6, 2000 at 5:31 pm

    On Wed, Apr 05, 2000 at 03:26:43PM -0400, John Tobey wrote:
    John Porter wrote:
    I wouldn't even mention it, except that mv'ing in CVS seems to be
    such a big deal: mv'ing a file in ClearCase is mindblowingly easy;
    it does the rmelem and mkelem automatically (of course, the source
    and destination dirs have to be checked out).
    OK, here's a band-aid:

    http://john-edwin-tobey.org/Perl/cvsmv

    PS We're all aware of VCS::CVS (and VCS::Cvs), yes?


    --

    Michael G Schwern http://www.pobox.com/~schwern/ schwern@pobox.com
    Q. What are "seven candles", "seven cows", and "seven heads of the beast
    which rises from the sea"? A: Metaphors. They are all symbolic. Which means,
    they are not real things.
    --Alex Chiu, Immortality Guy
  • Adam Turoff at Apr 5, 2000 at 7:01 pm

    Michael G Schwern wrote:
    PS Has anyone experience with Bitmover? Larry ummm... Rosler is it?
    was bubbling over about it at the Linux Expo '98 but I never got a
    chance to really look at it.
    Larry Rosler did the GRT (among other things).

    Larry McVoy does BitKeeper (www.bitmover.com).

    I'm pretty sure the GRT and BitKeeper are as mutually exclusive as
    these two Larrys. :-)

    -- Adam
  • Russ Allbery at Apr 5, 2000 at 8:04 pm

    Michael G Schwern writes:

    PS Has anyone experience with Bitmover? Larry ummm... Rosler is it?
    was bubbling over about it at the Linux Expo '98 but I never got a
    chance to really look at it.
    I fully expect BitKeeper to rock. The only question will be whether
    someone can beat into Larry's head that it can actually work as free
    software. (It's not worth having the argument until after he releases it,
    though; he's heard too many arguments about it at this point and isn't
    listening any more.)

    --
    Russ Allbery (rra@stanford.edu) <http://www.eyrie.org/~eagle/>
  • Chip Salzenberg at Apr 5, 2000 at 6:06 pm

    According to Bennett Todd:
    I think the implication that CVS is like Windows is pretty much pure
    flame-bait.
    Well, that's why I mentioned asbestos. :-)

    I REALLY don't like CVS's limitations. I've never considered it as a
    good system, because of those limitations. So my attitude toward it
    is, in those ways, similar to my attitude toward Windows.

    But to be more precise, CVS's limitations are like those of Windows
    _filesystems_. On the other hand, Windows isn't the only system with
    such limited file semantics. On the gripping hand, I couldn't have
    said "VMS", or no one would have gotten the point.
    --
    Chip Salzenberg - a.k.a. - <chip@valinux.com>
    "I wanted to play hopscotch with the impenetrable mystery of existence,
    but he stepped in a wormhole and had to go in early." // MST3K
  • Dan Sugalski at Apr 5, 2000 at 6:28 pm

    At 11:06 AM 4/5/00 -0700, Chip Salzenberg wrote:
    But to be more precise, CVS's limitations are like those of Windows
    _filesystems_. On the other hand, Windows isn't the only system with
    such limited file semantics. On the gripping hand, I couldn't have
    said "VMS", or no one would have gotten the point.
    Well, not no one. (I'm presuming, of course, that you're not conflating
    Windows and VMS' filesystems, given VMS doesn't have those silly
    limitation) Why, want a CMS library for this? I can drag the Vaxstation
    II/GPX (A slower system you've never seen. 1/3 blazing VUPS of processing
    power... :) out and attach it to the net, though there's still the issue of
    clients.

    Dan

    --------------------------------------"it's like this"-------------------
    Dan Sugalski even samurai
    dan@sidhe.org have teddy bears and even
    teddy bears get drunk
  • Chip Salzenberg at Apr 5, 2000 at 8:43 pm

    According to Dan Sugalski:
    VMS doesn't have those silly limitation.
    I could have sworn that Perl's test suite needed to be very careful to
    e.g. not rename a file while it's open under VMS... ?
    --
    Chip Salzenberg - a.k.a. - <chip@valinux.com>
    "I wanted to play hopscotch with the impenetrable mystery of existence,
    but he stepped in a wormhole and had to go in early." // MST3K
  • Dan Sugalski at Apr 5, 2000 at 8:58 pm

    At 01:42 PM 4/5/00 -0700, Chip Salzenberg wrote:
    According to Dan Sugalski:
    VMS doesn't have those silly limitation.
    I could have sworn that Perl's test suite needed to be very careful to
    e.g. not rename a file while it's open under VMS... ?
    Nope. Go ahead and do what you like to a file--rename, delete, mock
    incessantly--it works. Maybe you were thinking of the windows stuff? (We
    can open a file by our equivalent of an inode, *with* full access control,
    too. Nyah! :) You may need some jacked privileges depending on who owns the
    file and who's got it open, but the name and file are only losely coupled
    in VMS.

    It's not hugely common, mind, but I do it occasionally.

    Dan

    --------------------------------------"it's like this"-------------------
    Dan Sugalski even samurai
    dan@sidhe.org have teddy bears and even
    teddy bears get drunk
  • Chip Salzenberg at Apr 7, 2000 at 12:37 am

    According to Dan Sugalski:
    At 01:42 PM 4/5/00 -0700, Chip Salzenberg wrote:
    I could have sworn that Perl's test suite needed to be very careful to
    e.g. not rename a file while it's open under VMS... ?
    Nope. Go ahead and do what you like to a file--rename, delete, mock
    incessantly--it works. Maybe you were thinking of the windows stuff?
    I'm probably conflating VMS's weird pipe handling with the open/rename
    limitations of Windows and (now that I think of it) OS/2.
    --
    Chip Salzenberg - a.k.a. - <chip@valinux.com>
    "I wanted to play hopscotch with the impenetrable mystery of existence,
    but he stepped in a wormhole and had to go in early." // MST3K
  • Ilya Zakharevich at Apr 5, 2000 at 6:19 pm

    On Wed, Apr 05, 2000 at 12:20:40PM -0400, Bennett Todd wrote:
    But I really like its balance of complexity -vs- functionality; I
    like the way it exposes as much as you want of its underpinnings. By
    sticking strictly to a directory heirarchy of RCS files, it may end
    up with limitations that turn icko when you want to rename things
    (although create and remove isn't an awful way of handling it). But
    it retains lovely compatibility with other tools.
    Why sticking with RCS is a limitation? I can move RCS files around
    without any problem. All what is needed is a file with a mapping
    "real name" => "the name in the source tree".

    Ilya
  • Bennett Todd at Apr 5, 2000 at 6:26 pm

    2000-04-05-14:19:53 Ilya Zakharevich:
    Why sticking with RCS is a limitation?
    Not just sticking with RCS; sticking with RCS and attempting to
    ensure that the RCS tree comes as close as possible to representing
    the exact state of the real tree.
    I can move RCS files around without any problem. All what is
    needed is a file with a mapping "real name" => "the name in the
    source tree".
    Yup, if you're willing to lose the ability to just look at the RCS
    tree and nothing else and [nearly] reconstruct the represended
    directory heirarchy (modulo only deletions), then it's easy to add
    another data structure to allow you to track renames. Is it worth
    it? Opinions differ.

    For my tastes, CVS is ideal just the way it is. Don't add more
    complexity. If you really want to rename a file, delete the old one
    and add the new one. If you want to rearrange the entire directory
    heirarchy, just re-import the whole kit and kaboodle as a new tree.

    But different folks have different demands; Chip apparently prizes
    the ability to rename and rearrange regularly, to the point where
    this limitation in CVS totally pisses him off, so badly that he
    wants to warm his toes by a nice flamewar just to relax and
    unwind:-).

    Good thing we don't have a monopoly situation here, where anybody is
    gonna be forced to use CVS if they don't like it.

    -Bennett
  • Ilya Zakharevich at Apr 5, 2000 at 6:30 pm

    On Wed, Apr 05, 2000 at 02:25:53PM -0400, Bennett Todd wrote:
    I can move RCS files around without any problem. All what is
    needed is a file with a mapping "real name" => "the name in the
    source tree".
    Yup, if you're willing to lose the ability to just look at the RCS
    tree and nothing else and [nearly] reconstruct the represended
    directory heirarchy (modulo only deletions),
    Given that the file above lives in the RCS tree ;-), this is not a
    limitation.
    But different folks have different demands; Chip apparently prizes
    the ability to rename and rearrange regularly, to the point where
    this limitation in CVS totally pisses him off, so badly that he
    wants to warm his toes by a nice flamewar just to relax and
    unwind:-).
    Well, I would not mind if `diff -u' could handle moves of a large chunk
    of text from one place to another in a meaningful way. ;-)

    Ilya
  • Chip Salzenberg at Apr 5, 2000 at 8:46 pm

    According to Ilya Zakharevich:
    On Wed, Apr 05, 2000 at 02:25:53PM -0400, Bennett Todd wrote:
    Yup, if you're willing to lose the ability to just look at the RCS
    tree and nothing else and [nearly] reconstruct the represended
    directory heirarchy (modulo only deletions),
    Given that the file above lives in the RCS tree ;-), this is not a
    limitation.
    Quite. All that's needed is a map showing where the file _used_ to
    be. You still get to see the plain RCS files in their current
    locations. But I don't think CVS will ever get that ability, seeing
    as how the CVS developers haven't done it yet. Too bad.
    Well, I would not mind if `diff -u' could handle moves of a large
    chunk of text from one place to another in a meaningful way. ;-)
    A long time ago (mid-80s?) I had someone's version of diff that
    actually could notice and represent _moved_ lines. It was helpful
    from time to time. Want me to dig it up? I might need to find a
    working 5 1/4" floppy. :-,
    --
    Chip Salzenberg - a.k.a. - <chip@valinux.com>
    "I wanted to play hopscotch with the impenetrable mystery of existence,
    but he stepped in a wormhole and had to go in early." // MST3K
  • Ilya Zakharevich at Apr 5, 2000 at 8:59 pm

    On Wed, Apr 05, 2000 at 01:45:11PM -0700, Chip Salzenberg wrote:
    Well, I would not mind if `diff -u' could handle moves of a large
    chunk of text from one place to another in a meaningful way. ;-)
    A long time ago (mid-80s?) I had someone's version of diff that
    actually could notice and represent _moved_ lines. It was helpful
    from time to time. Want me to dig it up?
    I do not think it would be useful without matching functionality of
    `patch'. :-(
    I might need to find a working 5 1/4" floppy. :-,
    I have the same problem. One of my computers's 5 1/4" floppy is still
    working, but it is in Berkeley, and I spend a lot of time in Ohio. ;-)

    Ilya
  • John Porter at Apr 5, 2000 at 8:59 pm

    Chip Salzenberg wrote:

    A long time ago (mid-80s?) I had someone's version of diff that
    actually could notice and represent _moved_ lines. It was helpful
    from time to time. Want me to dig it up? I might need to find a
    working 5 1/4" floppy. :-,
    I seem to recall such a thing too.

    Hey -- a quick check of ClearCase's diff shows that it has this
    capability.

    --
    John Porter
  • John van V. at Apr 5, 2000 at 4:22 pm

    --- Joshua N Pritikin wrote:
    On Tue, Apr 04, 2000 at 09:03:51PM -0700, chip@valinux.com wrote:

    CVS is to ClearCase as Windows is to Unix.
    Odd coming from VA, writers of SourceForge
    >

    =====
    John van Vlaanderen

    #############################################
    # CXN, Inc. Contact: john@thinman.com # #
    # Proud Sponsor of Perl/Unix of NY #
    # http://www.thinman.com/puny #
    #############################################

    __________________________________________________
    Do You Yahoo!?
    Talk to your friends online with Yahoo! Messenger.
    http://im.yahoo.com
  • Chip Salzenberg at Apr 5, 2000 at 6:03 pm

    According to John van V.:
    CVS is to ClearCase as Windows is to Unix.
    Odd coming from VA, writers of SourceForge
    Hey, CVS is open source and does its job well, so my comparison wasn't
    entirely fair. It's just that CVS's self-chosen job is, IMO, too
    limited.

    But SourceForge is the product of a small (brilliant, dedicated) team
    within VA. If I'd been involved, we would at least be offering some
    alternative to CVS. I think.
    --
    Chip Salzenberg - a.k.a. - <chip@valinux.com>
    "I wanted to play hopscotch with the impenetrable mystery of existence,
    but he stepped in a wormhole and had to go in early." // MST3K
  • John van V. at Apr 5, 2000 at 6:18 pm

    Pretty much any really worthwhile file-move support, intended to be
    an improvement over delete/create...
    I'm going to implement SourceForge on my PUNY server pretty soon.

    Subsisting on a diet of perl and an little shell, I would really like to see
    method management along w/ the files. I have extractated my self from deep
    doodoo a number of times by renaming a method to _old and writing the fix in a
    few minutes.

    So, could we not encapsulate the best parts of CVS in XS, while adding admin
    funcitons (and method control) in perl until time permits C .= '++' coding.

    =====
    John van Vlaanderen

    #############################################
    # CXN, Inc. Contact: john@thinman.com # #
    # Proud Sponsor of Perl/Unix of NY #
    # http://www.thinman.com/puny #
    #############################################

    __________________________________________________
    Do You Yahoo!?
    Talk to your friends online with Yahoo! Messenger.
    http://im.yahoo.com
  • John van V. at Apr 6, 2000 at 2:15 pm

    Well, hm, if we could get a site license for SourceForge ... Hm,
    that's an interesting idea!
    I actually begged the perforce people to talk to the people at
    SourceForge. They seemed to have the impression that SourceForge wasn't
    willing to work with non-open source software. But why can't the
    developers decide? Chip, can't you do something?
    I was told at Linux Expo that sourceforge was downloadable.
    It that isn't so then I think I can safely predict it will be !!

    My info is that ny perl companies are really happy w/ CVS so I'm voting for a
    perl XS-ed release. Now all I have to do is get VA to move to NY-- trust me
    when I tell you IBM throws some really cool parties.

    It also looks like linux is getting a short form of a revival here, my friends
    sig, Linux Society, is already the talk in the bars. I'm, of course, shooting
    for a perlized boot/distro/admin engine.

    =====
    John van Vlaanderen

    #############################################
    # CXN, Inc. Contact: john@thinman.com # #
    # Proud Sponsor of Perl/Unix of NY #
    # http://www.thinman.com/puny #
    #############################################

    __________________________________________________
    Do You Yahoo!?
    Talk to your friends online with Yahoo! Messenger.
    http://im.yahoo.com
  • John van V. at Apr 7, 2000 at 2:09 pm

    I really agree with a position of using only free software..
    I have to agree. It may not be reasonable to enforce an all-free
    policy ... YET....

    ..but hey, it's making me think that we
    really need a good open source feature-a-like of ClearCase
    Gimp, as an example, may be the best thing that ever happened, spawing no less
    than 'n' psychedelic desktops. But I got a real shock yesterday when the text
    I got was telling me that I can't use it to draw or paint. Point being, dont
    limit yourselves, but then thats perl6's modular philosophy :)

    A perl6 specific CVS, with all the expansion hooks for the
    make/test/install/distro/config/update/educate features would go a long way to
    achieve a lot of admin and development goals.

    To me, the only un-fun part of this kind of a project would be pulling CVS
    apart to create the XS's. The re-construction phase, however, would be cake.

    =====
    John van Vlaanderen

    #############################################
    # CXN, Inc. Contact: john@thinman.com # #
    # Proud Sponsor of Perl/Unix of NY #
    # http://www.thinman.com/puny #
    #############################################

    __________________________________________________
    Do You Yahoo!?
    Talk to your friends online with Yahoo! Messenger.
    http://im.yahoo.com

Related Discussions