Re: pgk netlist

[David Bateman]

Bob Weigel wrote:
> - pkglist.m implements a package listing from internet.  
> - pkgget.m downloads a package.  
> - xurlwrite.m is used for backward compatability and to work around a problem 
> in urlwrite for Octave 2.9.9+ that has not been compiled with
> cURL (separate email on this issues sent to the maintainers list).
> - checkdeps.m is needed by xurlwrite.m
> 
> This a demonstration set of files that may be used as needed.
> 
> Here is a link to a related discussion 
> [http://www.nabble.com/Test-web-site-of-new-octave-forge-web-site-including-all-packages-tf2417959.html#a6748784].
>  
> 
> The following are my notes.  
> 
> Issues: 
> 
> * File integrity should be checked.  The easy way to do this is to have all 
> of 
> the files in a directory and have pkglist grab the file size from the 
> listing.  A more common method is to have the md5 listed in the same 
> directory of the file.

I'd thought to add an MD5 sum for each package to the download page, but ...

> 
> * Currently the package list is extracted from an html page in a
> delicate manner.  The "right" way to do it would be to store the Name,
> Description, URL strings in an XML file (say, info.xml) or a
> JavaScript Object in a javascript file, (say info.js).  Both the a js
> file in packages.html and the Octave program would then just need to be
> read from the info file.

if you want to have a specific file for the packages on the website that
is easier for you to read it would make more sense to add the MD5 sums
to this file instead of or as well as on the download page.


> * As written, downloading and parsing of the web page is not executed
> if packages.mat is found.  A better way to do this is to just look at
> the last-modified date of packages.html (from HTTP header) and
> re-download and parse if it has changes since packages.mat was
> created.

I had an issue with the idea of the pkgget command. Sourceforge requests
that all downloads go through a specific interface rather than directly
to one of the mirrors. They request this so that file download
statistics are correct. So this means that yes you will be able to go to
one of the mirror to get a package like

http://ovh.dl.sourceforge.net/sourceforge/octave/signal-1.0.0.tar.gz

but Sourceforge doesn't like you to do it. Until recently you couldn't
wget the package download page to get the package, but it appears they
have fixed this issue and so running wget on

http://downloads.sourceforge.net/octave/signal-1.0.0.tar.gz?download

will get the package..

Regards
David

> 
> Bob Weigel