bug#48543: Feature request: put ELN cache in $XDG_CACHE_HOME by default

bug#48543: Feature request: put ELN cache in $XDG_CACHE_HOME by default

[Clément Pit-Claudel]

Hi all,

By default the eln cache is stored in .emacs.d/eln-cache.  Would it make sense to store it in $XDG_CACHE_HOME/emacs/eln by default on platforms that obey the XDG spec?

I came across this because my backup scripts already exclude $XDG_CACHE_HOME, but didn't exclude .emacs.d/eln-cache, so I found myself backing up hundreds of megabytes of ELN files.

In a related Reddit discussion, https://www.reddit.com/r/emacs/comments/mltaie/gccemacs_how_to_permanently_change_where_elncache , Steve Purcell wrote:

> Now that emacs honours xdg settings for finding its own config, might it make sense for the eln cache to move to the xdg cache directory?

And Eli Zaretskii responded:

> I don't think so, personally: the XDG spec says the XDG cache directory should hold "user-specific non-essential (cached) data files". The *.eln files don't fit that description: the user would like the files to persist.
>
> That said, feel free to submit a feature-request bug report about this, and let's see what others think about this.

FWIW, I had a quick look through my .cache folder, and it seems that chromium and Firefox both put compiled code in that folder (JS bytecode and compiled Webassembly files).

bug#48543: Feature request: put ELN cache in $XDG_CACHE_HOME by default

[Lars Ingebrigtsen]

Clément Pit-Claudel <cpitclaudel@gmail.com> writes:

> By default the eln cache is stored in .emacs.d/eln-cache.  Would it
> make sense to store it in $XDG_CACHE_HOME/emacs/eln by default on
> platforms that obey the XDG spec?

I think that makes sense.  Perhaps Andrea has some comments; added to
the CCs.

> And Eli Zaretskii responded:
>
>> I don't think so, personally: the XDG spec says the XDG cache
>> directory should hold "user-specific non-essential (cached) data
>> files". The *.eln files don't fit that description: the user would
>> like the files to persist.
>>
>> That said, feel free to submit a feature-request bug report about
>> this, and let's see what others think about this.
>
> FWIW, I had a quick look through my .cache folder, and it seems that
> chromium and Firefox both put compiled code in that folder (JS
> bytecode and compiled Webassembly files).

Since the .eln files are cached data, it seems like the right place to
put them.  Of course the user would like all (non-stale) cached data to
persist.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#48543: Feature request: put ELN cache in $XDG_CACHE_HOME by default

[Eli Zaretskii]

> Cc: 48543@debbugs.gnu.org, Andrea Corallo <akrl@sdf.org>
> From: Lars Ingebrigtsen <larsi@gnus.org>
> Date: Wed, 13 Jul 2022 14:08:56 +0200
> 
> >> I don't think so, personally: the XDG spec says the XDG cache
> >> directory should hold "user-specific non-essential (cached) data
> >> files". The *.eln files don't fit that description: the user would
> >> like the files to persist.
> >>
> >> That said, feel free to submit a feature-request bug report about
> >> this, and let's see what others think about this.
> >
> > FWIW, I had a quick look through my .cache folder, and it seems that
> > chromium and Firefox both put compiled code in that folder (JS
> > bytecode and compiled Webassembly files).
> 
> Since the .eln files are cached data, it seems like the right place to
> put them.  Of course the user would like all (non-stale) cached data to
> persist.

So we prefer to hear more complaints about Emacs being slow to start
because the eln-cache was purged?  Don't forget that the XDG
directories can disappear without a warning.

bug#48543: Feature request: put ELN cache in $XDG_CACHE_HOME by default

[Lars Ingebrigtsen]

Eli Zaretskii <eliz@gnu.org> writes:

>> > FWIW, I had a quick look through my .cache folder, and it seems that
>> > chromium and Firefox both put compiled code in that folder (JS
>> > bytecode and compiled Webassembly files).
>> 
>> Since the .eln files are cached data, it seems like the right place to
>> put them.  Of course the user would like all (non-stale) cached data to
>> persist.
>
> So we prefer to hear more complaints about Emacs being slow to start
> because the eln-cache was purged?  Don't forget that the XDG
> directories can disappear without a warning.

This doesn't seem to worry the Chromium etc people, so I don't think it
should worry us.  

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#48543: Feature request: put ELN cache in $XDG_CACHE_HOME by default

[Visuwesh]

[புதன் ஜூலை 13, 2022] Lars Ingebrigtsen wrote:

> Eli Zaretskii <eliz@gnu.org> writes:
>
>>> > FWIW, I had a quick look through my .cache folder, and it seems that
>>> > chromium and Firefox both put compiled code in that folder (JS
>>> > bytecode and compiled Webassembly files).
>>> 
>>> Since the .eln files are cached data, it seems like the right place to
>>> put them.  Of course the user would like all (non-stale) cached data to
>>> persist.
>>
>> So we prefer to hear more complaints about Emacs being slow to start
>> because the eln-cache was purged?  Don't forget that the XDG
>> directories can disappear without a warning.
>
> This doesn't seem to worry the Chromium etc people, so I don't think it
> should worry us.  

Why not use XDG_DATA_HOME in that case?  It doesn't get cleared I think.

bug#48543: Feature request: put ELN cache in $XDG_CACHE_HOME by default

[Andrea Corallo]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> Eli Zaretskii <eliz@gnu.org> writes:
>
>>> > FWIW, I had a quick look through my .cache folder, and it seems that
>>> > chromium and Firefox both put compiled code in that folder (JS
>>> > bytecode and compiled Webassembly files).
>>> 
>>> Since the .eln files are cached data, it seems like the right place to
>>> put them.  Of course the user would like all (non-stale) cached data to
>>> persist.
>>
>> So we prefer to hear more complaints about Emacs being slow to start
>> because the eln-cache was purged?  Don't forget that the XDG
>> directories can disappear without a warning.
>
> This doesn't seem to worry the Chromium etc people, so I don't think it
> should worry us.  

Hi Lars,

thanks for adding me in Cc.

Browser speed is tipically dominated by the internet/web server
bandwidth and it's therfore very variable.  Users will hardly notice if
a page loads slower because the cache was removed and complain, this is
indeed not the case for Emacs as program.

Also given the XDG spec definition I'm not sure this move would be the
right thing.

Best Regards

  Andrea

bug#48543: Feature request: put ELN cache in $XDG_CACHE_HOME by default

[Stefan Kangas]

Visuwesh <visuweshm@gmail.com> writes:

> Why not use XDG_DATA_HOME in that case?

I realize that the XDG Base Directory Specification could be more clear
on these details, but I believe that a study of the current practices as
well as previous discussions on e.g. the XDG mailing list, in Debian,
Arch and so on, would lead to the conclusion that this would be wrong.

It is generally understood that XDG_DATA_HOME is roughly analogus to
/usr/share and XDG_CACHE_HOME is roughly analogous to /var/cache -- one
is volatile, the other one is not.

For starters, please compare the contents of ~/.local/share with
~/.cache on your machine to see what other software puts where.

Next, to understand how these issues are often approached, I invite you
to study references such as:

    https://wiki.debian.org/XDGBaseDirectorySpecification
    https://wiki.archlinux.org/title/XDG_Base_Directory
    https://wiki.gnome.org/action/show/Initiatives/GnomeGoals/XDGConfigFolders
    https://bugzilla.mozilla.org/show_bug.cgi?id=259356

There are more to find obviously, but that should be plenty to get the
discussion started.

> It doesn't get cleared I think.

AFAIK, $XDG_CACHE_HOME also doesn't "get cleared" arbitrarily.  There
has to be some action on behalf of the user for that to happen.