This is part of The Pile, a partial archive of some open source mailing lists and newsgroups.
Date: Wed, 20 Mar 2002 15:47:48 +0200 (IST) From: Ira Abramov <lists-svlug@ira.abramov.org> To: make bzlilo <svlug@svlug.org> Subject: [svlug] CVS or better? looking around I see Subversion isn't ready yet (called itself pre-alpha), bitkeeper are non-free(beer) for non-free(speach), and so I'm stuck again with CVS by default... or does anyone know something I don't about subversion or some other CMS? CVS just seems like a very problematic system to keep using for non-OSS development... basically, we're a young startup with no huge cash reserves, and we need a free-but-good CMS for linux, any ideas will be appreciated. === Date: Wed, 20 Mar 2002 06:27:26 -0800 (PST) From: "J. Paul Reed" <preed@sigkill.com> To: Ira Abramov <lists-svlug@ira.abramov.org> Subject: Re: [svlug] CVS or better? On Wed, 20 Mar 2002, Ira Abramov wrote: > CVS just seems like a very problematic system to keep using for non-OSS > development... How so, specifically? I know of tons of non-OSS (Cisco, Netscape, Sun, etc.) companies who use it for their bread and butter version management, and have no trouble with it (or put another way, have worked around CVS's quirks either through hacking tools to do what they need, i.e. CVSweb/LXR/Bonzai/etc. or by implementing a specific set of processes/policies within their development process). As I remember, Cisco did, however, switch to something else for their IOS version management, but it *was* CVS for the longest time. > basically, we're a young startup with no huge cash reserves, and we need > a free-but-good CMS for linux, any ideas will be appreciated. Mmm... I think you want CVS. ;-) === Date: Wed, 20 Mar 2002 00:19:25 -0800 From: Karen Shaeffer <shaeffer@neuralscape.com> To: svlug@svlug.org Subject: Re: [svlug] CVS or better? On Wed, Mar 20, 2002 at 03:47:48PM +0200, Ira Abramov wrote: > > I'm stuck again with CVS by default... > Ira Abramov I think you ought to generate a list of things your group needs to do--but cannot seem to get accomplished using CVS. Then you would have the basis to make informed, objective decisions about CVS and or other systems. === Date: Wed, 20 Mar 2002 18:14:32 +0200 (IST) From: Ira Abramov <lists-svlug@ira.abramov.org> Cc: make bzlilo <svlug@svlug.org> Subject: Re: [svlug] CVS or better? On Wed, 20 Mar 2002, J. Paul Reed wrote: > > CVS just seems like a very problematic system to keep using for non-OSS > > development... > > How so, specifically? the long list of "problems with CVS" is well known and you can find its different versions all over the net. start with the homepage of subversion :) > > basically, we're a young startup with no huge cash reserves, and we need > > a free-but-good CMS for linux, any ideas will be appreciated. > > Mmm... I think you want CVS. ;-) I hate making decisions by default... when there are no second choices, trouble begin. I'll spare you the standard freedom-of-choice speach that follows. === Date: Wed, 20 Mar 2002 18:18:29 +0200 (IST) From: Ira Abramov <lists-svlug@ira.abramov.org> Cc: make bzlilo <svlug@svlug.org> Subject: Re: [svlug] CVS or better? On Wed, 20 Mar 2002, Karen Shaeffer wrote: > I think you ought to generate a list of things your group needs to do--but > cannot seem to get accomplished using CVS. Then you would have the basis to > make informed, objective decisions about CVS and or other systems. I agree that in normal market situ with multiple choices competing, that would be the right thing to do. in our case, since there are 3 or less competitors (only one?) you take what you can get, and the way is to list the pros and cons the 2-3 products have in comparison to eachother and do the math. if I can't even find a single contender on CVS' title, there's no competition to be drawn. === Date: Wed, 20 Mar 2002 08:54:34 -0800 (PST) From: "J. Paul Reed" <preed@sigkill.com> To: Ira Abramov <lists-svlug@ira.abramov.org> Cc: make bzlilo <svlug@svlug.org> Subject: Re: [svlug] CVS or better? On Wed, 20 Mar 2002, Ira Abramov wrote: > I agree that in normal market situ with multiple choices competing, that > would be the right thing to do. It's interesting to talk about this particular product in terms of a market... CVS is so standard and ubiquitous, that there really *is* no market for (free as in beer and freedom) version control software. But unlike Microsoft, there's a reason no market for version control software exists: CVS has done the job quite nicely for thousands (millions?) of developers for at least a decade. It's everywhere. All open source developers you hire will understand it. It doesn't break. A lot of the problems subversion's webpages mention really are problems... and some of them have fixes that come in the form of scripts or other addons; www.cvshome.org has a lot of these. > if I can't even find a single contender on CVS' title, there's no > competition to be drawn. The fact that there *aren't* contenders to CVS (at least, serious ones right now) should tell you something... namely that CVS is rock solid, won't corrupt your data/code, and most of the complaints about it have workarounds (either technical or procedural... and a lot of the procedural ones have arguments for why you should be doing your development "that way" anyway). Having said that, I glanced at the webpages, and subversion looks cool; I'll be keeping my eye on the project as it progresses. But I (especially) wouldn't trust my startup's intellectual property to it... yet. === Date: Wed, 20 Mar 2002 10:45:09 -0800 From: J C Lawrence <claw@kanga.nu> To: svlug@svlug.org Subject: Re: [svlug] CVS or better? On Wed, 20 Mar 2002 15:47:48 +0200 (IST) Ira Abramov <lists-svlug@ira.abramov.org> wrote: > bitkeeper are non-free(beer) for non-free(speach)... Correction: BitKeeper is free (as in beer) as long as you are willing to publish changeset comments. It is also free as in speech as long as any source changes you make do not cause the results to fail the regression tests. === Date: Wed, 20 Mar 2002 20:47:31 +0200 (IST) From: Ira Abramov <lists-svlug@ira.abramov.org> Cc: make bzlilo <svlug@svlug.org> Subject: Re: [svlug] CVS or better? On Wed, 20 Mar 2002, J. Paul Reed wrote: > On Wed, 20 Mar 2002, Ira Abramov wrote: > > > I agree that in normal market situ with multiple choices competing, that > > would be the right thing to do. > > It's interesting to talk about this particular product in terms of a > market... CVS is so standard and ubiquitous, that there really *is* no > market for (free as in beer and freedom) version control software. > of course there is a market, as people ARE using CVS. did you mean there is no need for a new product in this market? Most everyone I ask says they use it whilst grinding their teeth and missing the ClearCase/Perforce/whatever they were using at that other workplace a year ago, but it's "good enough" so they make do. there are some solid good reasons for subversion to be developped, the same basic reasons sendmail is not enough and Exim/qmail/postfix are taking over its market - patch-upon-patch software that is mainly used since it was the first of breed and is ubiquitous by default. > But unlike Microsoft, there's a reason no market for version control > software exists: CVS has done the job quite nicely for thousands so does microsoft office, according to the same logic. why would anyone need 1-2-3 or Wordperfect? Am I on a linux list here? > and some of them have fixes that come in the form of scripts or other > addons; www.cvshome.org has a lot of these. patch over patch, like I said. Add a third party management interface, then two helper scripts, and suddenly your workarounds conflict with the way the management thingy likes to do things, etc. === Date: Wed, 20 Mar 2002 10:53:49 -0800 From: J C Lawrence <claw@kanga.nu> To: make bzlilo <svlug@svlug.org> Subject: Re: [svlug] CVS or better? On Wed, 20 Mar 2002 08:54:34 -0800 (PST) J Paul Reed <preed@sigkill.com> wrote: > The fact that there *aren't* contenders to CVS (at least, serious ones > right now) should tell you something... namely that CVS is rock solid, > won't corrupt your data/code, and most of the complaints about it have > workarounds (either technical or procedural... and a lot of the > procedural ones have arguments for why you should be doing your > development "that way" anyway). Sadly, none of these facts are true. CVS is not rock solid, is prone to data corruption when used over NFS (and has no veracity checking tools), is prone to creating broken repositories, has numerous basic flaws which cannot be effectively worked around, and with extremely rare exception the "work-arounds" made for CVS are just that: excuses for shoddy software. The flaws, weaknesses, and stupidities of CVS are well known. Don't disregard them. Recognise them for what they are: bugs and (usually uninformed) trade-offs. If you can live with those bugs and trade-offs, fine, but I would hardly advocate that for many projects. === Date: Wed, 20 Mar 2002 18:21:04 -0800 (PST) From: "J. Paul Reed" <preed@sigkill.com> To: <svlug@svlug.org> Subject: Re: [svlug] CVS or better? On Wed, 20 Mar 2002, J C Lawrence wrote: > Sadly, none of these facts are true. CVS is not rock solid, is prone to > data corruption when used over NFS (and has no veracity checking tools), Huh. I've used it (at work and at home) over NFS for years now, without incident. That, of course, doesn't prove that it's not prone to data corruption... it's simply one data point. Granted, those NFS mounts were across 2 switches/hubs in the same data closet at most... if you run NFS through 37 routers across the country (like Cisco does), you get what you deserve, since you're not using NFS what it was designed for. > The flaws, weaknesses, and stupidities of CVS are well known. Don't > disregard them. I'm not. CVS can be very stupid and crufty at times, but to paraphrase: "CVS is the worst version control system; except for all others." I'm just saying that it's ubituitous and it's a standard, and it works relatively well for what it was designed to do, and it has years of doing what it does under its belt. I *ALSO* said (which everyone ignored) that Subversion looked really cool... and obviously there's a need for it if CVS is so utterly bad (although, if CVS is so utterly bad, I've never run into all of its pitfalls... maybe I do froofy development or something). BUT, I wouldn't trust (especially) my startup's code to some product that's not even considered by its authors to be alpha yet. === Date: Wed, 20 Mar 2002 18:42:42 -0800 (PST) From: "J. Paul Reed" <preed@sigkill.com> To: make bzlilo <svlug@svlug.org> Subject: Re: [svlug] CVS or better? On Wed, 20 Mar 2002, Ira Abramov wrote: > of course there is a market, as people ARE using CVS. did you mean there > is no need for a new product in this market? I think I meant there's really no competition in the market because of CVS. Blech. It's finals week here... I don't know what I meant. ;-) > Most everyone I ask says they use it whilst grinding their teeth and > missing the ClearCase/Perforce/whatever they were using at that other > workplace a year ago, but it's "good enough" so they make do. Yeah, but I would argue those are apples and oranges, and outside the scope of your original question. If you'd like to spend your startup's money on ClearCase/Bitkeeper/whatever, then that changes the name of the game, and the type of suggestions you'll get. But cheap/free was part of your initial requirements. > so does microsoft office, according to the same logic. No, not at all. Office is a monopoly because Microsoft abused its power in the OS market to extend its monopoly to the productivity market. > why would anyone need 1-2-3 or Wordperfect? They *are not* the same thing. You pay for Office. Or Word Perfect. Or 1-2-3. You don't pay for CVS. There are companies pushing Office, Word Perfect, and 1-2-3, trying to get people to buy them; there's no company pushing developers to use CVS. The rules of markets (I admit, I misused the term), market segements, marketing, and general economics aren't the same in the face of open source software. Don't try to compare them. > Am I on a linux list here? Sometimes I wonder. > and suddenly your workarounds conflict with the way the management thingy > likes to do things, etc. What? That sentence doesn't parse. Anyway, that would be a valid argument if we were talking about closed-source software. CVS is open source; you can change the way the "management thingy" likes to do things so it's how you like to do things. That's the whole point. === Date: Wed, 20 Mar 2002 18:57:48 -0800 From: J C Lawrence <claw@kanga.nu> To: make bzlilo <svlug@svlug.org> Subject: Re: [svlug] CVS or better? On Wed, 20 Mar 2002 18:21:04 -0800 (PST) J Paul Reed <preed@sigkill.com> wrote: > On Wed, 20 Mar 2002, J C Lawrence wrote: >> Sadly, none of these facts are true. CVS is not rock solid, is prone >> to data corruption when used over NFS (and has no veracity checking >> tools), > Huh. I've used it (at work and at home) over NFS for years now, > without incident. NFS is stateless. Ergo, NFS doesn't support locking. LockD is poorly supported, has frequent comparability problems across platforms, is unenforced at the system level, has non-guaranteed semantics, and thus merely lends the illusion of NFS being anything other than a virus. > That, of course, doesn't prove that it's not prone to data > corruption... it's simply one data point. However, given the popularity of NFS and given the fact of widespread far reaching corruption of RCS based repositories (like CVS) due to silent NFS corruption (largely due to the afore mentioned locking failures) one would think that any reasonable CMS system would implement internal protections (such as the much superior SCCS). Aside: Large chunks of many older versions of commercial Unixes are no longer retrievable due to the simple fact of silent NFS corruption of RCS .v files. More noticeable is that the only OS vendor in the Valley to use a non-RCS based tool (Sun and TeamWare) has no such problems -- a fact which is explicitly due to TeamWare being SCCS based and SCCS being based on forward rather than reverse diffs. CVS doesn't even acknowledge the problem. > Granted, those NFS mounts were across 2 switches/hubs in the same data > closet at most... if you run NFS through 37 routers across the country > (like Cisco does), you get what you deserve, since you're not using > NFS what it was designed for. NFS was mostly designed for shared read-only access, or minimally single-writer access. >> The flaws, weaknesses, and stupidities of CVS are well known. Don't >> disregard them. > I'm not. CVS can be very stupid and crufty at times, but to > paraphrase: "CVS is the worst version control system; except for all > others." Strongly disagreed. I'd quibble as to where exactly CVS belongs in the bottom of the pile, and don't really care about its exact placement either, but its a pretty shitty product and there are far better products and implementations out there. > I'm just saying that it's ubituitous and it's a standard, and it works > relatively well for what it was designed to do, and it has years of > doing what it does under its belt. Nope. You should talk to Richard Kingdon (CVS author) about CVS (I did). It is ubiquitous, it is a defacto standard within the OSS community, it is poorly design, poorly implemented, mostly fails to fulfill its initial design criteria, but has successfully ossified into a largely uninspected dinosaur lump that sites secure on its throne solely for the reason that it got there first. > I *ALSO* said (which everyone ignored) that Subversion looked really > cool... and obviously there's a need for it if CVS is so utterly bad > (although, if CVS is so utterly bad, I've never run into all of its > pitfalls... maybe I do froofy development or something). I've previously detailed CVS' shortcomings to this and other lists several times. Rather than repeat them here again ad nauseum, please check the archives. > BUT, I wouldn't trust (especially) my startup's code to some product > that's not even considered by its authors to be alpha yet. I generally consider Subversion to be uninteresting, even if/when it leaves Alpha, but that's due more for the fact that they're not actually doing anything interesting or useful beyond attempting to fix a few of the more glaring oversights in CVS. === Date: Wed, 20 Mar 2002 19:28:35 -0800 (PST) From: "J. Paul Reed" <preed@sigkill.com> To: make bzlilo <svlug@svlug.org> Subject: Re: [svlug] CVS or better? On Wed, 20 Mar 2002, J C Lawrence wrote: > NFS is stateless. So? CVS doesn't use lockf(), flock(), fcntl(), or any other "file lock" calls that have to be implmeneted with lockd; its locking is done via mkdir(), which is atomic even over NFS. If you don't believe me, search the info-cvs list archives over at gnu.org. And I will point out: the fact that you didn't know this should seriously cause you to question everything you thought you "knew" about CVS and what makes it so bad... 'cause I sure am. > > Granted, those NFS mounts were across 2 switches/hubs in the same data > > closet at most... if you run NFS through 37 routers across the country > > (like Cisco does), you get what you deserve, since you're not using > > NFS what it was designed for. > > NFS was mostly designed for shared read-only access, or minimally > single-writer access. As an aside, everyone complains about NFS, and even I have banged my head against the wall with it a couple of times. But I must admit, if you take time and care when setting it up, and set it up across switches or hubs (i.e. not 45 routers), with NFS implementations from the same vendor, it *does* work. Of course, it's not as cool as Sun's original vision of NFS, I'm sure, but it's not the "virus" you make it out to be, IF it's implemented/used with some common sense, *where* it makes sense to do so. > Strongly disagreed. Great. No need to discuss it then. It should be noted that you've failed to convince me that CVS is all that bad. When you take into account the requirements of free, standard, and stable (I'm not convinced that it's not yet, because no one has shown me anything where there's evidence of people losing tons of data), I'd recommend and use CVS. A huge portion of the OSS community, it seems, agrees with me, even if it's because their reason is "It sucks, but there's nothing better." If that's not good enough for you, Ira,, for whatever reason, please report back to SVLUG in 6 months to a year on what you used instead, and why you used it... I'm sure we'd all (myself included) be interested. I'm done talking about this; I need to go check something into my NFS-mounted repository. === Date: Wed, 20 Mar 2002 19:44:09 -0800 From: J C Lawrence <claw@kanga.nu> To: make bzlilo <svlug@svlug.org> Subject: Re: [svlug] CVS or better? On Wed, 20 Mar 2002 19:28:35 -0800 (PST) J Paul Reed <preed@sigkill.com> wrote: > On Wed, 20 Mar 2002, J C Lawrence wrote: >> NFS is stateless. > So? > CVS doesn't use lockf(), flock(), fcntl(), or any other "file lock" > calls that have to be implmeneted with lockd; its locking is done via > mkdir(), which is atomic even over NFS. Actually the ___only___ operation which is guaranteed atomic over NFS is creat(2), and mkdir() is only guaranteed atomic to the degree that it requires a creat() within its syscall (true for all current FS implementations I know of but not guaranteed), but that's another matter. More simply, CVS locking, while recently improved, is not consistent (tho I haven't checked for a couple years now so it may have improved). Historically I've found it comparatively easy to corrupt CVS repositories because of this and have done so repeatedly. That all said, as a user my biggest beefs against CVS are lack of a changeset model, pathetic 3-way merge support, lack of meta data tools, and non-existent rename support. >> NFS was mostly designed for shared read-only access, or minimally >> single-writer access. ... > But I must admit, if you take time and care when setting it up, and > set it up across switches or hubs (i.e. not 45 routers), with NFS > implementations from the same vendor, it *does* work. Of course, it's > not as cool as Sun's original vision of NFS, I'm sure, but it's not > the "virus" you make it out to be, IF it's implemented/used with some > common sense, *where* it makes sense to do so. Agreed, for the case detailed above. I just don't find that a very interesting or particularly useful case and there are generally better tools to use for that case (eg HTTP). > No need to discuss it then. It should be noted that you've failed to > convince me that CVS is all that bad. I've not stated it was all bad. Few to no things are. I do consider it to be generally Bad Software and easy to avoid/replace with other significantly preferable tools, but those are different things. As always its a set of value choices. Some people happen to like Trabats. > If that's not good enough for you, Ira,, for whatever reason, please > report back to SVLUG in 6 months to a year on what you used instead, > and why you used it... I'm sure we'd all (myself included) be > interested. FWLIW I've been using BitKeeper almost exclusively for the last few years under their OpenLogging license (ie no $$$) and have been and remain absurdly pleased with it. === Date: Wed, 20 Mar 2002 22:40:21 -0800 (PST) From: Robert Hajime Lanning <lanning@lanning.cc> To: claw@kanga.nu (J C Lawrence) Cc: preed@sigkill.com (J. Paul Reed), Subject: Re: [svlug] CVS or better? CVS can also work across the network in other ways than NFS. NFS is just "A Bad Thing"(tm) from locking to security. (ok "L" through "S" is not much of the alphabet.) :) You can use CVS through rsh (bad security) or ssh (good security) or even use their "pserver" method (plain text password, but different than login password. Also, this is how I use it to store sendmail binaries, ".cf" files and other supporting files for twelve mailrouters.) === Date: Wed, 20 Mar 2002 23:21:24 -0800 From: J C Lawrence <claw@kanga.nu> To: svlug@svlug.org Subject: Re: [svlug] CVS or better? On Thu, 21 Mar 2002 08:39:39 +0200 (IST) Ira Abramov <lists-svlug@ira.abramov.org> wrote: > On Wed, 20 Mar 2002, J. Paul Reed wrote: >> You don't pay for CVS. > well, that's why I am not putting koffice on the same level as Lotus > Office, nor CVS with Perforce. I was talking about comparing at the > same price level, and nothing competes with CVS. it's the ONLY public > server offered by Sourceforge and its equivilants as well AFAIK... BitMover will willingly host repositories for OpenSource projects using BitKeeper, and IIRC Perforce has (occasionally?) been persuaded to do likewise. I've also heard of a couple other shops offering free BitKeeper hosting, but I failed to keep track of them unfortunately. === Date: Thu, 21 Mar 2002 01:14:40 -0800 From: J C Lawrence <claw@kanga.nu> To: Karen Shaeffer <shaeffer@neuralscape.com> Cc: svlug@svlug.org Subject: Re: [svlug] CVS or better? On Wed, 20 Mar 2002 12:32:35 -0800 Karen Shaeffer <shaeffer@neuralscape.com> wrote: > On Wed, Mar 20, 2002 at 07:44:09PM -0800, J C Lawrence wrote: > I knew you would get around to your plug for bitkeeper. Hahahaha!!!! Hehn. I tried to stay away from an advocacy rant. Its too old a horse. > Hey JC, please explain to us about the pros and cons of bitkeeper. I think you've read my previous rants on the area, si I won't cover too much old ground there. What I haven't commented on previously are the specific features and capabilities of BK that I really like, that impress me, or which evoke the thought, "Damn, I'm glad BK does this!" when I use them. Such things enclude: -- A proper changeset model. CVS users tend to just not understand why this is such a significant item, or make comments about developer discipline and best practices. Poppycock. Changesets regularly both save my arse, even on single developer projects, and make manipulating and working with repositories easier. Good stuff. Not unique to BK to be sure (cf ClearCase, Perforce, etc), but nice. I'd probably love BK if the only change it made over CVS were merely adding proper changeset/transaction model (Subversion is doing most of but not all of this). The fact that BK takes the changeset/transaction thought and follows it through all the way to its logical conclusion (and much of what comes below is a reflection of this single point) is possibly BK's single largest value add. In truth, that is BK's base value: it is a designed product sitting on meticulously developed and proven math, not a set of marketing derived we-can-sell-this-feature bits (Larry has very good mathematicians on staff for that purpose alone). -- All repositories are first class repositories. I love this. Essentially it means that I can check out a repository from another repository and can check out new copies etc from the one I just made. There's no enforced master/child relationships. Build any sort of distribution pattern you want, tree shaped or ad-hoc/random. It all works. A small thing, but a killer thing. I maintain largely duplicate repositories for the same source bases (eg my entire $HOME is a BK repository with all my bash scripts, .xemacs configs, browser bookmarks, galeon configs etc) with copies on more than a score machines currently. There is no canonical "master repository" and it just doesn't matter. As I'm on each machine I can add scripts, tune configs, etc, check them in, and then later push or pull them off onto other machines as/when they're needed and BK just takes care of it all and makes it work. I just don't have to care -- it just works. Heck, I'm almost in the position where I'm going to check all/most of my Debian installation into BK and use BK rather than APT-GET for file distribution. That way I could get the native advantages of a CMS for all my config files as well as a full history of all package changes. -- 3 way merging. BK's new pro-merge tool is brilliant. Better than brilliant. Its one the of the very few development tools I know of for which I can't think of a functional change that would actually make it better (actually I can't think of any others). Gobsmacking. -- Meta data. CVS doesn't try. Perforce handles it fairly well. ClearCase goes hog wild here. Meta data is the sort of thing you don't tend to notice or use until its there, but once it is you start using it and relying on it. As such its another thing that the CVS folks tend to look at you funny about as they don't see the values. -- Reproducibility. Not unique to BK by a long shot, but good stuff and something I rely on regularly. I can delete files, rename files, rename and delete directories, make changes, revert changes etc and know for absolute certain that if I check out a version of the repository as of a past date that I'll get a bit-image accurate copy as of that time, with every deleted file put back, every renamed file put back, etc, bit image accurate as of the time of the revision I checked out. If I'm working with someone else and I get a changeset from them, I know that my repository (as of that changeset) is bit-image identical to their repository. If I check out as of that rev I can guarantee a bit image identical repository to what they have sitting in front of them -- which helps debugging a lot. -- Nested repositories. A small thing, but nice. Its good to be able to aggregate a project out of BK repositories arranged in a natural directory hierarchy. Makes life easier. -- I like the GUI. I'm not much of a GUI guy, but I like BK's GUI. There's not much of it, and many commands and much functionality doesn't have any GUI, but the bits that have GUis benefit from that and actually add value from having a GUI. Its a small thing, but I like BK's citool -- it solves real problems and makes my CMS life better. > I realize it is well written software, but what are the pros and cons > about OpenLogging? OpenLogging in itself I don't consider that much of an advantage. I don't mind, even mildly like the idea, but it also doesn't affect me much. Some of my stuff is logged, some not (single developer exception). Its cute to go check the site and see some of log comments but it really has no direct value for me. Were something like SourceForge implemented under BitKeeper (it would be trivially easy to copy/import all of SF's CVS repositories into BK), or BK to get the sort of mass OSS exposure that CVS has, then OpenLogging would be a lot more attractive -- of a sudden you would have the simple ability to find out who is working on what, where, and how much for the entire planetary OSS market. Very nice. Very useful. Doesn't exist now. That said, we're starting to get that with the Linux Kernel and the advantages there are already clearly visible. === Date: Fri, 22 Mar 2002 11:02:03 -0800 From: "Ian MacLure" <imaclure@seagull.com> To: svlug@svlug.org Subject: [svlug] gCVS usability ( or lack thereof ) I'm doing some pathfinding for converting a project to a Linux environment and one aspect of this is, of course, CM. We are actually supposed to be using Rational's tools ( inter-alia ClearCase ) but we have some network capacity issues that aren't going away anytime soon. So, for now, we are falling back upon CVS which has served well during a proof of concept phase which was done with that alternative OS from Redmond. Anyhoo, the CVS GUI we used was WinCVS and were quite pleased with it. I'd like to try and leverage familiarity with WinCVS for our Linux conversion and of course gCVS immediately presents itself. All fine and good except it doesn't work worth beans. On a RedHat 7.2 system ( several on diverse h/w sets ) pointed at a SunOS server ( Ultra-5 as it happens ), there are a bunch of problems starting with login being broken. gCVS is quite happy to browse a local s/w tree created using TkCVS but has problems doing much else. gCVS installed cleanly. Anybody have any experience with gCVS particularly "gotchas" I should be aware of? ============================================================ === Date: 22 Mar 2002 11:38:17 -0800 From: Sean Allen <zeroone@worldonline.co.za> To: svlug@svlug.org Subject: Re: [svlug] gCVS usability ( or lack thereof ) On Fri, 2002-03-22 at 11:02, Ian MacLure wrote: <snip/> > Anybody have any experience with gCVS particularly > "gotchas" I should be aware of? No, unfortunately. However, if your development environment is *emacs based, I cannot speak highly enough of pcl-cvs for emacs. (anonftp from ftp.lysator.liu.se/pub/emacs) === Date: Fri, 22 Mar 2002 15:31:34 -0800 (PST) From: Roland Krause <rokrau@yahoo.com> To: svlug@svlug.org Subject: [svlug] Re: gCVS usability ( or lack thereof ) Ian, I hope this is useful to you allthough it isnt directly related to gCVS. > Subject: [svlug] gCVS usability ( or lack thereof ) I did a very similar thing here at work and I use cervisia (http://cervisia.sf.net/) which I have on Solaris x86 and Sparc as well as on Linux. Cervisia is absolutely stable and almost complete in terms of functionality. I have my windows side covered by WinCVS also and I have briefly looked into a WinCVS lookalike called LinCVS but found no reason for a switch. Cervisia has a fantastic 'merge window' in which resolve conflicts from merges, it has repository browse possibility support for tags, branches, edits and watches... you name it. Best of all, it can run as a konqueror part if you are into that sort of thing. Best regards, Roland PS: ClearCase still beats CVS in terms of functionality but ClearCase's unix GUI is severly outdated, buttugly and seemingly unmaintained and therefore even less an option for me. === Date: Fri, 22 Mar 2002 04:27:02 +1030 (CST) From: Richard Sharpe <rsharpe@ns.aus.com> To: Ira Abramov <lists-svlug@ira.abramov.org> Cc: make bzlilo <svlug@svlug.org> Subject: Re: [svlug] CVS or better? On Thu, 21 Mar 2002, Ira Abramov wrote: > On Wed, 20 Mar 2002, J. Paul Reed wrote: > [Deletia] > > There are companies pushing Office, Word Perfect, and 1-2-3, trying to get > > people to buy them; there's no company pushing developers to use CVS. > > there's something worse working here, a default choice for lack of > competition. basically if you start your way into professional > programming in the Academic world or the OSS community you will find > yourself rushed by everyone to use CVS "because everybody uses it" and > little other reason. that's why I was reminded of M$ office: here in > Israel we never got Hebrew support from WordPerfect or Lotus, and > Microsoft took over the tiny market (Dagesh, Qtext, Aleph-Bet and > Einstein were all excellent wordprocessors in the 80s and early 90s), > first because it was graphic and worked in windows, but then because > "that's what everybody used". the day offices had to send out documents > in source form (Email) rather than print, M$Office became the standard > and all other wordprocs died within 2-3 years. > > So in a way, the Internet, in the most democratic way possible, is > creating and keeping software monopolies alive, be they browsers or CMS. > Sadly, in a non-Darwinian way, it's not always the best-of-breed that > triumphs, but some aligator that hasn't evolved in 150 million years > because it's good enough. While I may have mis-understood your intent here, I think that the use of the term 'non-Darwinian' in the context of 'best-of-breed' is an incorrect portrayal of what Darwin was getting at and what Evolution is all about. The 'eye' that most animals have is not 'best-of-breed'. Cephalopods demonstrate that, but the eye structure you and I have has a near-monopoly. So, I see strong parallels. Good enough is, indeed, good enough. ===