New Port Version Check feature

[Fernan Aguero]

+----[ Edwin Groothuis <edwin at mavetju.org> (04.Oct.2005 21:54):
|
| After a long weekend of coding (long live the Queen!) this morning
| I have sent out this morning the first batch of emails to alert
| maintainers of possible availability of newer versions of the ports
| they maintain.
| 
| The full overview of the first run can be found at
|     http://edwin.adsl.barnet.com.au/~edwin/ports/
| 
| It works by replacing the current port version with "logical" next
| versions: For foo-1.2.3 it tries foo-1.2.4, foo-1.3.0 and foo-2.0.0.

Edwin, 

let me thank you for this terrific idea. It's always been a
pain to look for new versions of ports. I can rest now that
you've automated this :)

One thing I noticed: In the email I received from you, 2 out
of 6 cases are cases of your script finding a new distfile
that is not (yet) announced in the main pages. Or one that
is a testing (candidate) release that was not labeled as
such in the distfile name (but otherwise clearly stated in
the home page).  

Even though I would not say in these cases that 'a new
version' is out, I would not try to filter or fix your
script, since this is very useful to have for those cases
when you're not following closely the development of the
port.

| Possible ceavats:
| 
| - Some mastersites have been added to the ignore list. For example
|   GForce websites have the habbit of returning a file which doesn't
|   exist. And ftp://cr.yp.to/ has the habbit of being able to chdir
|   to a file which doesn't exist, thus totally confusing LWP.
| 
| - Webservers which return the content-type text/* get ignored. Why?
|   Because of too many false positives from websites which don't let
|   their 404 handler return the 404 status code.
| 
| - It handles source/1.2.3/foo-1.2.3, but it doesn't handle
|   source/1.2/foo-1.2.3 fully (thanks to tmclaugh@#bsdports for
|   mentioning)
| 
| - Ports with version numbers in the name can give false positives:
|   lang/gcc32, lang/gcc33 say that there is gcc-4.0.0 available
|   (thanks to gerald@ for mentioning)

[ ... ]

| Edwin
|
+----]

I would add:

 - Ports without version numbers in the name can give false
   negatives: biology/seaview is one example from the ports
   I maintain. 

In this case, downloading the distfile and checking for a
change in the md5 sum would find a discrepancy. But this is
already detected and reported by the automated build
scripts.

And (but perhaps you're already taking care of these): ports
in which a date is used in replacement of a version number
can also be difficult to track. 

Finally, I guess that to help you make your script better we
all would have to manually check the ports we maintain and
report if there are ports with new versions that were missed
by your script. 

Thanks again!!!

Fernan