I just read brian d foy's article on the dilemma of those who provide software to others. I am not sure about this part though:
Most programmers, no matter the language, focus on the least time-consuming portion---the actual coding.
If coding is the least time-consuming part how long should people be kept waiting for a first release? I guess if Lincoln D. Stein were to rewrite CGI.pm from scratch plus doing all the other work (like providing documentation, creating facilities for easy installation etc....the stuff that is allegedly more time consuming) a release wouldn't be due within the next half year. And then on releasing a distribution he could still not be sure to have anticipated all problems that users could be confronted with. This part is actually impossible as I realize ever so often with surprise.
I tend to see Perl modules as a colaborative effort: the author is just the one making a suggestion to a large peer-group (the first release) and within time his ideas and the needs of the users start to converge. The earlier this convergence starts the better. I think this is also true (in a way) for any open source software.
Now as for time-consumption: I happen to observe that from one release to the next there is often jsut a difference of a few changed lines in code. The actual work comes after that: a preliminary distribution needs to be uploaded to several different machines to be tested. If you have ever worked over a slow ssh-lines on machines with broken terminals and no usable text-editor you may guess how much this takes. Then count how often you boot-switch from, say, Linux to Windows and back again on your local machine till all things appear to work ok. Add to that the time and hassle it needs to create a ppm distribution and so on and so forth. This is all done in order to provide the utmost convenience to users.
I am aware that I am focusing on modules whereas brian has talked about software in general. I think a Perl module and a ready-to-run application have different target-groups. The first appeals to people being familiar (at least a little) with programming: They are thus used to solving unforeseen problems. End-users often aren't. In parts we are already taking this into account: You instinctively wouldn't release a command-line application for Win32 whereas nothing is more common on Linux.
As always, there doesn't seem to be a trivial solution to that. When it comes to complaining about lack of convenience, the Perl community is sitting on a rather high horse (fortunately). Things aren't perfect but at least we know very well that they are far worse anywhere else.
