applications backend data

Luigi Toscano luigi.toscano at
Tue Mar 13 14:53:34 UTC 2018

Aleix Pol ha scritto:
> On Tue, Mar 13, 2018 at 11:45 AM, Harald Sitter <sitter at> wrote:
>> On Tue, Mar 13, 2018 at 12:05 AM, Albert Astals Cid <aacid at> wrote:
>>> El divendres, 9 de març de 2018, a les 14:08:44 CET, Harald Sitter va
>>> escriure:
>>>> quick fyi: I am working on a change for the backend data of applications
>>>> new tech is meant to automatically crawl appstream appdata out of CI
>>>> builds. eventually, I'd like that to be an api service you can talk to
>>>> for app metadata (for now it's still on-disk blobs though).
>>>> the data is mostly similar to the json blobs we have right now, albeit
>>>> different keys (also the blobs contain localization data)
>>>> the overall format is outlined here (except we dump json instead of yaml):
>>>> please also keep this in mind for the new website. so long as the
>>>> frontend bits are sufficiently abstract from the backing data we
>>>> should be fine though.
>>>> I hope to get this done today so it's easier to know what I am talking
>>>> about.
>>> Can we please next time get some heads up before you commit things on
>> Yup I've messed up there. Sorry.
>>> It's the second time in a row you commit things to and break it and it
>>> lowers my morale a lot.
>>> The idea of autogenerating files is good, it's how the original json files
>>> where generated, sadly your commit has caused regressions.
>>> There's dozens of broken links now, like for example, second link (for me) at
>>> As maintainer of rsibreak i find it to be an unacceptable regression.
>>> Also two months ago we agreed on this very list to still show unmaintained
>>> applications (with an unmaintained marker) for the same very reason of not
>>> breaking links.
>>> So what we need is a applications/ subfolder that at least contains the same
>>> data it contained before your change, if it contains less data for any
>>> application, that is a regression and even if it does show better data for
>>> other apps I do not feel it's an acceptable loss, even more without having had
>>> any discussion about it with the people that kept that data up to date.
>> I presently don't have the energy. I've reverted back to where things were.
>> HS
> How can we iterate this in a way that we don't get the worse of both
> worlds? I'd say getting the information metadata from appstream is
> where we should be, if it's a matter of introducing some redirects,
> maybe it's easier to just do that, no?

Sure, that's the way but without regressions. Maybe testing it on a staging
website could help.


More information about the kde-www mailing list