RE: [Condor-devel] Installation issues.

[Bruce Beckles <mbb10@xxxxxxxxx>]

On Thu, 2 Jun 2005, Alain Roy wrote:

<snip>
> > > Don't - we're taking condor_install out of the distribution soon.
> >
> > That's too bad - it worked flawlessly.  In particular, it made
> > modifications to the global config file (not just the local config file)
> > during the installation process.  IMHO this is really, really important.
>
> We want to make condor_configure do the same thing.
>
> condor_install confuses people: it asks tons of questions that people don't 
> know how to answer.

Just to add my 2 cents worth: we really like condor_install and have found 
condor_configure impossible to use.

I've lost track of the number of installations of Condor I've had to go 
and fix because someone misused condor_configure, whilst I've never had to 
do that with condor_install (possibly because the defaults are fairly 
sensible).

I've found that with a very small amount of guidance people can use 
condor_install correctly, and what's more, all the descriptive text it 
provides gives them some handle on what is going on.

condor_configure, by contrast, has been found difficult to use by everyone 
over here who's tried it, and one of the key points in feedback as to why 
this is so is that they have *no* idea what it is doing or why it is doing 
it.  In addition, its error messages are confusing and less than helpful.

I have tried giving people condor_configure and the relevant section of 
the manual (and no guidance form myself) and also similarly with 
condor_install.  The results have always been that those using 
condor_install were fine, those using condor_configure had problems 
(sometimes minor problems, but problems nonetheless).

I think you are making a mistake to get rid of condor_install, and I think 
it will make the life of people who have to support other people's Condor 
installations (like I do) more difficult.

I think that decisions about the "user interface" of an installation 
process of a product should always be made on the basis of user trials and 
user feedback, and, crucially, validated by further user trials. 
Obviously I believe from my own limited user trials that you are either 
not doing this or doing this on a set of users very different from mine 
(which I would find odd).

	-- Bruce