[cvsnt] Proposed CVSNT Setup - Looking for Feedback

[Torsten Martinsen]

Dianne Chen wrote:

>1) Is the server spec'd appropriately for the intended
>usage/community? I was unable to find any
>configuration recommendations.
>  
>
Sounds very reasonable. Where I work, we have around 40 people using a 
CVS server of similar specification, which also handles other stuff like 
Samba and Bugzilla. Most of the stuff in CVS is source code, but quite a 
lot of it is Microsoft Office documents (i.e. binary files).

>2) Does the community have any recommendations on
>flavor of Linux? Does the community have first-hand
>knowledge of what to use vs. what to avoid, what took
>more work to get running, etc.
>  
>
Debian (testing) works for me. There is no (AFAIK) Debian package of 
CVSNT, but apt-get makes everything else easy to handle.

>3) For a reasonable first-pass at tools, does the
>proposed set  look usable/complete? What developer
>VC-activities are not covered by the above tools?
>  
>
I would choose TortoiseCVS as a client (but then I'm bound to say that 
:-). You should put ViewCVS on the server as well.

Some people prefer VC to be integrated into their IDE - if your 
favourite IDE supports the SCC API, you can buy a third-party plugin 
(cheaply) at http://pushok.com/.

>4) Of specific concern, can I propose a Linux
>Server/Winxp client split as shown above and have
>those tools play together well?
>  
>
Yes, no problem. We have recently switched to using the SSPI protocol, 
which avoids a lot of complaints when people's Windows password expires.

>5) Am I correct that one should limit the activities
>occurring on the CVS server? Does that mean it should
>*only* do cvs functions and nothing more? What have
>others "gotten away with" doing on their cvs servers
>and at what cost?
>  
>
As hinted above, you don't have to go to extremes. But you should 
probably not allow e.g. shell access on the CVS server.

>6) How do others handle builds? Do all builds occur on
>client machines? Do some make use of build servers? In
>my previous experiences with Clearcase, we used CM
>servers AND build servers, but those were for some
>pretty big products.
>  
>
Developers build on their own machines, the weekly/nightly build runs on 
a dedicated build server.

>7) If builds occur on client machines, does that mean
>in order to use "make", everyone has Cygwin installed
>on all of their PCs?
>  
>
Well, basically, if you want to run "make", you must have "make" :-) One 
way of getting a decent make is to install Cygwin (which I find is also 
very valuable for the odd script to do builds, extract metrics from 
Bugzilla, etc. as opposed to the horror that is cmd.exe - but I 
digress), but many development environments already include their own make.

-Torsten