Provide package-size sort option in package lists

[TJ]

I realise I could manually work through the HTTP listings for the PPA to figure out what files are using the most space in the PPA, in order to determine what, if anything, could be deleted.

However, there should be an easier way to do that. Something as simple as displaying a package-size column that allows ordering.

[MMS-Prodeia (mms-prodeia)]

I would really appreciate this, as of 25th of february PPA's get limited to 1GB ... so having an easy way of knowing how large ones PPA is would help a lot!

[Celso Providelo (cprov)]

The repository-size section is active again in edge.l.n. You can quickly see the total size of files in you repository and that's the value used for checking quotas.

Regarding the original bug report, I don't really see users browsing the package list for package size. Maybe it's just me, but being small or big doesn't make a package less or more important.

Can we come up with real use-cases involving PPA cleanups ?

[TJ (tj)]

Celso, the use-case is very clear, as MMS-Prodeia stated.

I recently got the email warning about PPA limits from the 25th too. My PPA summary shows it uses 1.5GB but I have *no* way to analyse which packages are using the most space in order to make judgements as to which might be removed, or whether to ask for an increase in size.

My packages usually contain proposed bug-fixes and so can expire once upstream or the Ubuntu repositories include the patches. However, many of the patches won't get upstream any time soon or may not be appropriate for SRU so I need to keep them in the archive if the improvements are going to remain available to Ubuntu users.

The only way to determine the package source/binary sizes is to manually inspect the archive listings for each package - a very time-consuming and frustrating process!

The issue is compounded by the fact that the Delete functionality only lists the first 20 or so packages and so I found myself in the ridiculous situation of having to filter by each letter of the alphabet in turn to review all the packages in the archive whilst trying to figure out which might be deleted.

At least for the Delete functionality, there should be a column displaying the package's binary and source sizes since it is and will become an important criteria as time goes on and PPAs maintain older packages whilst continually adding newer versions.

[Celso Providelo (cprov)]

TJ,

All good points, I just don't get how package-size might affect your decision for keeping or deleting a PPA package.

On the other hand, this the relevance of the source changelog and package-diff is very clear, and for that you need to inspect the details for each publication anyway.

That said, I see other informations that we could provide off-hand (w/o requiring you to expand the package row) that are way more important than the package size itself, for instance, installabilty against ubuntu (can this package be installed in ordinary ubuntu systems, release + security + updates)

This factor clearly affects quotas, those packages are useless in the repository and can be deleted straight away (unless you have special clients).

I don't have any problem in summing up the size of the files in a package within the expandable area as "Repository size: 300 MiB" but listing that in the package rows isn't in the interest of most users accessing the PPA page.

The package deletion interface has problems related to its limited batch and search capabilities and the fact that selections can not yet be preserved if we use batching, but that's not necessarily related with package-sizes being available or not, IMO.

The central point is this bug is if you have a dozen of packages and your PPA is over-quota, if they are still relevant in your context you need more space, doesn't matter how big each of them are.

[TJ (tj)]

On Fri, 2009-02-13 at 21:03 +0000, Celso Providelo wrote:
>
> All good points, I just don't get how package-size might affect your
> decision for keeping or deleting a PPA package.

I see it as a very important aspect that informs on regular
house-keeping. For example, I recently provided OpenJDK6 with a patch
that fixes a serious bug causing applets to fail to load. Although I
posted the fix upstream it only prompted a discussion and, so far as I
am aware, the only way to get it is from my PPA.

The package took up a large amount of space in the PPA, both source and
binary (which is usually 3 x package size with the architecture-specific
binaries) so on it's own it was responsible for more than 350MB.

I had to balance that against how many people seemed to be making use of
it (here's where download statistics would be *very* helpful). I could
only judge by the lack of significant posts to the Ubuntu forums where I
announced availability.

Often, I build packages for multiple releases to provide back-ports that
wouldn't otherwise be available to LTS and other releases.

I also had builds of Evolution (to fix a task-list bug that causes a
seg-fault), Xorg components (to enable XDMX), and other eaters of space.

All of them were suffering from up-stream paralysis and/or lack of
attention/decision by the relevant Ubuntu team and not suitable for SRU.

At the rate I bug-fix my PPA could quickly consume 10GB and more if
every package I create were uploaded, so in my scenario at least, the
more help I can get in determining popularity and size in order to
decide what is most important to keep, the better.

I quickly abandoned my plans to host bug-fix test packages of the kernel
based purely on the size constraints.

> That said, I see other informations that we could provide off-hand (w/o
> requiring you to expand the package row) that are way more important
> than the package size itself, for instance, installabilty against ubuntu
> (can this package be installed in ordinary ubuntu systems, release +
> security + updates)

Could the size information be made available via the API so we could at
least distribute a common PPA maintainer package that would allow
developers to streamline ongoing PPA package management?

> I don't have any problem in summing up the size of the files in a
> package within the expandable area as "Repository size: 300 MiB" but
> listing that in the package rows isn't in the interest of most users
> accessing the PPA page.

I was suggesting it only for the 'Delete' page where the decision is
going to be made.