Re: URL for forge packages

[Carnë Draug]

On 16 August 2012 13:14, Carnë Draug <address@hidden> wrote:
> Hi
>
> the current method to get the URL for packages when installing with
> the -forge flag is to parse the HTML of the package for the version
> and use it to generate a URL to download it. I see two problems with
> this:
>
> 1. if we are to change the format of the documentation page this will
> start failing
> 2. if we move the packages to another host (as has been discussed
> before), it will also fail
>
> While such changes are not planned to happen on OF soon, when they do
> it will mean that users using octave versions before the change won't
> be able to use the -forge flag. I think to prevent this before becomes
> an issie it's a good idea.  I would like to propose to have a simple
> text file on octave.org with this info. 1 line with package names and
> their URLs would suffice. This would also make the code to get
> packages from forge simpler since the parsing would also be simpler.
>
> Any comments?

I gave this some more thought and here's the solution I propose. The
idea will be to have a text file in octave.org that is automatically
updated with a file I'll set up in SF. The following syntax for such
file would allow for other projects similar to Octave Forge (if and
when they appear) to also have their -project flag in pkg.

=Forge http://path-text-file-with-list-of-forge-packages
package1-name package1-version package1-url
package2-name package2-version package2-url
package3-name package3-version package3-url
=ProjectX http://path-text-file-with-list-of-projectx-packages
package1-name package1-version package1-url

Basically, the word after the equal would be the project and flag used
to get packages, followed by a URL for a text file that is
responsibility of that project. That text file would have just the
project, version, and url lines. Should be pretty simple for a perl
script to routinely check those URL's for changes and edit the file on
octave.org.

I'd guess that XML would be better but having to parse them in Octave
is more complicated and I don't think we need that.

Any comments? Can I start writing such scripts?

Carnë