ClearCase label issue

Some people prefer ClearCase, but this is probably due to a lack of understanding how ClearCase works. I herewith summarize some issues I had with ClearCase:
– Renaming a file, renames the file in all historical versions of the file…. you better take a copy of your ClearCase VOB before thinking about renaming.
– Labeling a file, and moving the previous label. There is no clue where the previous label was, nor who did it. You better take a daily archive of your complete structure as a back-up (and store this archive on a real system).
– Fixing the read-only mark of a file. You like the file getting orphaned or eclipsed…. this is a good start.
– Adding a file to a folder in two different development tracks. Good luck, you will be tempted to believe that your file should be present in all environments, versions, …
– Creating a version of a folder where there are active check-outs, well, they will be included in all versions of a file.
– Finding out if there are files check-out, not in ClearCase, … or whatever in +200 folders… Good luck, I suggest you move to a unix environment first, and write some script.
– Changing the casing of a file. Good luck, it just crashes your Vob, your client, your view, … You can try again but it will fail always. But you know, you can keep 2 versions in the Vob of the same file, if your casing is different.
Label a set of files at a certain moment in the future. Impossible without advance scripting.

The Cause:
Because ClearCase mounts VOBs the same way for all platforms, and tries to resemble a FAT file system, it just does use the lowest denominator of all, and uses the read-only mark of a file to act upon. (At least SVN stores this information in another place.)
Because ClearCase does not support group-commit. It cannot handle the directory at the same moment it handles the file.
Because ClearCase is evil.

The Solution:
-SVN
-Git
-Any other system

Source control statistics

People are using following systems:

CVS 12,6%
Git/GitHub 6,8%
IBM Rational ClearCase 2,8%
IBM Rational Team Concert 0,9%
IBM Rational Telelogic Synergy Suite 0,2%
Mercurial 3,0%
Microsoft Team Foundation Server 0,8%
Microsoft Visual SourceSafe 1,6%
Perforce 3,0%
Subversion 58,3%
None – I don’t use a source code management system 5,6%
Don’t know 1,4%
Other (specify) 2,9%

Source: Eclipse Community Survey 2010

Deploy svn on your server

SVN deployments are not always easy and straightforward. I prefer to have a web-interface such as WebSVN that will allow you to see the source code and check differences in between versions from a simple web-client. This is fine for reading, but not useful for getting a working copy to your local machine.

For downloading the source-tree, the branches or trunk, one would use one of the access protocols. The most straightforward is HTTP. This is understood by everyone and allows a simple viewer even without WebSVN. For commits, we would use ideally HTTPS. Both protocols require the WebDAV protocol to be deployed.

If you are working on a local machine, or in a local Unix/Linux/Mac environment, one would choose SVN or SVN+SSH. Given the difficulty of setting up SSH in a corporate environment from a Windows machine, I do not believe this is a real option for a vast majority of users. (Yes, some developers are living in a Windows corporate world.)

Lets assume the following scenario. Developer is somewhere on the internet, the repository is in corporate walls and a firewall is sitting in between. The correct set-up would be. HTTPS on the developer’s pc, connect to proxy forwarder gateway at corporate DMZ, forward this to in-house HTTPS WebDAV enabled SVN server.

The trouble with this set-up is that while the developer can browse the repository over HTTPS (or HTTP) or by means of the WebSVN directory, he cannot check-in if his pc is located behind a Microsoft proxy server…

Error getting OPTIONS – error 503, or MKACTIVITY – error 503.

I am still looking into a resolution for this problem. Any help welcome….