simplestreams json or index.asc for deliverables in the new list-of-deliverables?

[Vincent Batts]

On 19/05/15 09:05 -0400, Fabian Deutsch wrote:
>----- Original Message -----
>> On Mon, 18.05.15 19:44, Matthew Miller (mattdm at fedoraproject.org) wrote:
>>
>> > FESCo Ticket #1427 <https://fedorahosted.org/fesco/ticket/1427>
>> > formalizes the list of deliverables for each Fedora release. Cool.
>> >
>> > Let's go a step beyond this and make a machine-readable list, at least
>> > of image-based deliverables. One approach would be to use using Ubuntu's
>> > simplestreams
>> > format, documented at
>> > <http://bazaar.launchpad.net/~smoser/simplestreams/trunk/view/head:/doc/README.>
>> >
>> > A second option would be the index.asc as used by virt-builder:
>> > <https://fedorahosted.org/rel-eng/ticket/5805>
>> >
>> > That way, we'd have a standardized way for applications to find and
>> > download/launch/whatever various Fedora images. This is a definite need
>> > (see https://fedorahosted.org/cloud/ticket/93) and rather than
>> > inventing a new wheel, maybe we could see if we could pound one of
>> > these into shape.
>>
>> There's also the ACI discovery stuff:
>>
>> https://github.com/appc/spec/blob/master/SPEC.md#app-container-image-discovery
>
>If you are just referring to this part of the spec …
>
>> Doesn't really apply to non-ACI images right now, but the concepts in
>> it are a bit nicer I think then the alternatives (i.e. the fact that
>> it renders the image names into URL via a fixed template, and that it
>> uses <meta> tags for exposing this information is pretty nice).
>
>… then I agree. The URL scheme based identifier is nice.
>
>But it is lacking stuff like a description and the size of the image.
>And the method to retrieve metadata informations is somewhat cumbersome.

Description, maybe. But since it's just an HTML meta tag, then the page
itself could be self serving of that information. 

Size, it's just a flat file. An HTTP HEAD, an Content-Length header is
typically sufficient.

As for the retrieval, I can't say it's entirely cumbersome. Parsing HTML
is pretty straight forward. Serving this kind of information will
eventually be usable as well, as Docker is looking to accommodate
similar.
https://github.com/docker/distribution/pull/303/files#diff-4846d2808cb787c146c1c82ab2a69fdbR135
Though in that case, the serving a flat ACI files is cleaner and nicer.

>My 2ct.
>
>- fabian
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 181 bytes
Desc: not available
URL: <http://lists.fedoraproject.org/pipermail/devel/attachments/20150519/ab851856/attachment.sig>