Clearcase (was: Re: Information Required)

Thu Jan 15 16:54:11 UTC 2015

TL;DR: a lot of Clear-Case bashing below. Please skip if you are not
interested in that sort of thing.

On Thu, Jan 15, 2015 at 12:12 PM, Dirk Heinrichs <dhs at recommind.com> wrote:

>  Am 15.01.2015 um 11:49 schrieb Angel Ezquerra:
>
>  Unfortunately I've had to use Clear-Case for many, many years. In fact
> at some point I even made a small custom clear-case GUI to complement the
> one that was provided by Rational. Over the years I've used a pretty
> vanilla Clear-Case installation, another with some custom scripting layer
> on top to force a particular development workflow and another based on the
> UCM workflow. They all were really terrible.
>
> Agreed for UCM. Tried to avoid it whereever possible :)
>
>   It is actually hard to choose where to start when criticizing
> Clear-Case. There are so many wrong things about it! From its slowness,
>
> If CC is slow, your setup is wrong. I've worked in really fast CC
> environments.
>

If CC is slow perhaps you did not spend a ridiculous amount of money in
infrastructure and support, or you did and even then it is still slow.
Also, I question what you call "really fast". Does it compare to the
performance of any decent DVCS?

   to its out-dated model,
>
>
> It's different, but I won't call it outdated. In fact, I prefer its
> per-element checkout as it prevents accidental changes better.
>

It is definitely outdated IMHO. All popular SCMs developed in the last 10
years (or more) have moved away from the per-element checkout and into a
changeset based revision model. And god forbid you forget to checkout a
file that you want to modify. If you don't you get exposed to the concept
of hijacked files. If that is not outdated I don't know what is.

   to its incredibly clunky windows client,
>
> Matter of taste, I guess.
>

I never saw anyone praise the clear-case client in any way. Maybe I can no
longer say that?

The Clear-Case GUI looks old and ugly and has not seen a significant change
in years (possibly in decades). Doing anything remotely complicated through
it is most often an exercise in frustration (e.g. creating a new view for
instance takes an incredible amount of clicks, or dealing with hijacked
files, ugh!, or comparing changesets (if you use UCM) or any non single
file comparision is a pain, etc).

   to how hard it makes it to go back to older versions of your source code,
>
> Huh? Change a label in your configspec, ready.
>

Not if you use UCM. Also, you may be able to go back, but branching from
there is not that easy (it may even be impossible depinding on your server
configuration).

   to the brittleness of config-specs,
>
> Can you elaborate some more? We've always used some standard 5-line
> templates, quite easy.
>

You must version control your config spec if you want to be able to go back
to what you had at a given point in time. But then, that is why new SCMs do
that for you by tracking changesets instead of single files. Even if you do
that, depending on how complicated your config spec is, or if you use a lot
of modules things can get out of hand very easily.

   to the lack of any advanced history editing features,
>
> Only git has those, anyway. In CC, one can at least change the comments.
> That's more than you can do with plain Mercurial.
>

Only git? You said that with a straight face on the mercurial mailing
list?! ;-)

   to the huge need of administration infrastructure and overhead...
>
> Agreed, already stated this in my previous mail. However, you completely
> omitted its pros:
>

I don't see (m)any, that's why.

> Direct filesystem access to all versions via MVFS.
>

Using a good mercurial GUI client (e.g. TortoiseHg) you get the same
benefit without the weirdness or the slowness. Plus you get file annotation.

> Consistant filesystem paths (/vob/<vobtag>/...)
>

I would call that forced directory structure, which is a huge pain.

> Reproducible, distributed, parallel builds (clearmake)
>

I've not used clearmake much, but there are very good commercial and open
source tools that can do that sort of thing, which is not really an SCM
feature, IMHO.

   I am actually very surprised. You are the very first person that I've
> ever known to prefer Clear-Case to mercurial or git (and which had used
> these newer systems).
>
> I have been working as a certified CC admin for a long time ;)
>

That is probably my nightmare job ;-)

I guess that I made it clear by now that I think that CC is just horrible.
That does not mean that there is nothing good about it. The command line
client is OK, I guess (for what it does). It compares favorably with the
TFS command line client, for example (for what that is worth). It pays the
bills for a lot of unfortunate souls who work as cercified CC admins... ;->

If you work for a very process heavy company, where there is a
control-freak attitude towards who can do what with your code base, then
perhaps Clear-Case ifs for you. If you appreciate the sanity of your
developers, not so much IMHO :->

Cheers,

Angel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mercurial-scm.org/pipermail/mercurial/attachments/20150115/a14de94a/attachment-0002.html>