The toughest part of
it will be to ensure that the dependencies of a module matches another
module in PyPI.
I don't see it this way. As far as I'm concerned, you just have to say "The
dependencies are X, Y and Z. I can download and install X and Z but you'll
have to find Y for yourself." Enough irate (or motivated) users later, and
someone will upload Y to PyPI :)
From a viewpoint of a human user, yes, that can be expected. We will in
fact expect the person to google for it if it is not in PyPI. From a
viewpoint of a tool, that may be too tricky.
Another thing in my radar is that package names may change. I was
reading an article on O'Reilly ONLamp.com, "Using PySOAP" by Cameron
Laird, written on 06/14/2001
(www.onlamp.com/pub/a/python/2001/06/14/pysoap.html). When reading it, I
realise that instead of PySOAP, it should be SOAPpy. Of course, I do not
know if SOAPpy was named as PySOAP back then, but you see my point......
We can't hold things up just because PyPI isn't complete - that's going to
take quite a while to happen. For now we just have to assume that we can't
satisfy all dependencies by downloading and installing automatically.
begin:vcard
fn:Maurice Ling
n:Ling;Maurice
org:The University of Melbourne;Department of Zoology
adr:;;Gate 12, Genetics Lane;Parkville;Victoria;3010;Australia
email;internet:[EMAIL PROTECTED]
title:Probatory Ph.D. Candidate
tel;cell:+61 4 22781753
x-mozilla-html:FALSE
url:http://www.geocities.com/beldin79/
version:2.1
end:vcard
_______________________________________________
Catalog-sig mailing list
[email protected]
http://mail.python.org/mailman/listinfo/catalog-sig
