Compressing the Lisp and Info files in the MS-Windows installation

Compressing the Lisp and Info files in the MS-Windows installation

[Eli Zaretskii]

If the 'gzip' program is available, the Posix installation procedure
("make install") by default compresses all the *.el files in the lisp/
directory and its subdirectories of the installation tree, and also
compresses all the *info files in the info/ directory there.  Are
there any reasons not to do the same when Emacs is installed on
MS-Windows?

A related issue is whether the precompiled Windows binaries should be
distributed with these files compressed.  If they are compressed, end
users will need to have 'gzip' installed to be able to use them.

A compressed tree is 47 MB leaner than an uncompressed one (143 MB vs
190 MB).

I assume that people who prepare the binary distributions for Windows
will run something like "make install prefix=x:/foo/bar", and then
produce a zip file from the tree created at x:/foo/bar.  Thus the
above question is related both to local installations and preparing a
binary distribution for uploading to the FTP site.

(Even if the default is to compress, one can install without
compressing by saying "make install GZIP_PROG=".  So the question
really is what should be the default.)

Comments and opinions are welcome.

Re: Compressing the Lisp and Info files in the MS-Windows installation

[Stefan Monnier]

> A related issue is whether the precompiled Windows binaries should be
> distributed with these files compressed.  If they are compressed, end
> users will need to have 'gzip' installed to be able to use them.

AFAIK that's the main issue.  Can we expect the vast majority of
Emacs-on-Windows users to have gzip installed?
AFAIK the working hypothesis until now was "no".


        Stefan

RE: Compressing the Lisp and Info files in the MS-Windows installation

[Drew Adams]

> > A related issue is whether the precompiled Windows binaries 
> > should be distributed with these files compressed.  If they are 
> > compressed, end users will need to have 'gzip' installed to
> > be able to use them.
> 
> AFAIK that's the main issue.  Can we expect the vast majority of
> Emacs-on-Windows users to have gzip installed?
> AFAIK the working hypothesis until now was "no".

And even if they have gzip, some users will likely want to unzip all of the .el
so that they can more easily grep etc.  Is the gain from compression worth
making them go through that extra step?

Even though Emacs is bigger now than in the past, I would think that size
(download & disk) is generally less of a problem nowadays than it used to be.

Re: Compressing the Lisp and Info files in the MS-Windows installation

[Eli Zaretskii]

> From: "Drew Adams" <drew.adams@oracle.com>
> Cc: <emacs-devel@gnu.org>
> Date: Mon, 15 Apr 2013 13:13:14 -0700
> 
> And even if they have gzip, some users will likely want to unzip all of the .el
> so that they can more easily grep etc.

How is this different from users on Posix platforms?

In any case, uncompressing all of the files is easy on any platform
that has gzip installed.

> Is the gain from compression worth making them go through that extra
> step?

I don't know the answer to that.  I asked this question (and stated
the disk space savings) to hear opinions.  Hearing questions instead
of answers doesn't really help ;-)

> Even though Emacs is bigger now than in the past, I would think that size
> (download & disk) is generally less of a problem nowadays than it used to be.

Indeed.  But still on Posix platforms, compression is the default for
some reasons, although these same considerations clearly apply there
as well.

RE: Compressing the Lisp and Info files in the MS-Windows installation

[Drew Adams]

> > Even though Emacs is bigger now than in the past, I would 
> > think that size (download & disk) is generally less of a
> > problem nowadays than it used to be.
> 
> Indeed.  But still on Posix platforms, compression is the default
> for some reasons,
  ^^^^^^^^^^^^^^^^
> although these same considerations clearly apply there as well.

So the question is clarified from "Should the files be compressed on Windows
also?" to "What should the behavior be on Windows and (what should the behavior
be on) Posix (for Emacs Dev distributions)?".

Helper questions could be "Why is compression the default on Posix?" and "Why is
the _default_ Posix behavior important in deciding whether Emacs Dev compresses
or not for its distribution?"

Re: Compressing the Lisp and Info files in the MS-Windows installation

[Steinar Bang]

>>>>> Eli Zaretskii <eliz@gnu.org>:
>> From: "Drew Adams" <drew.adams@oracle.com>

>> And even if they have gzip, some users will likely want to unzip all
>> of the .el so that they can more easily grep etc.

> How is this different from users on Posix platforms?

It isn't.

(Why is why I don't think it is such a hot idea there either, but YMMV...)

> In any case, uncompressing all of the files is easy on any platform
> that has gzip installed.

>> Is the gain from compression worth making them go through that extra
>> step?

> I don't know the answer to that.  I asked this question (and stated
> the disk space savings) to hear opinions.  Hearing questions instead
> of answers doesn't really help ;-)

>> Even though Emacs is bigger now than in the past, I would think that size
>> (download & disk) is generally less of a problem nowadays than it used to be.

> Indeed.  But still on Posix platforms, compression is the default for
> some reasons, although these same considerations clearly apply there
> as well.

Re: Compressing the Lisp and Info files in the MS-Windows installation

[Eli Zaretskii]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Cc: emacs-devel@gnu.org
> Date: Mon, 15 Apr 2013 16:02:34 -0400
> 
> > A related issue is whether the precompiled Windows binaries should be
> > distributed with these files compressed.  If they are compressed, end
> > users will need to have 'gzip' installed to be able to use them.
> 
> AFAIK that's the main issue.  Can we expect the vast majority of
> Emacs-on-Windows users to have gzip installed?
> AFAIK the working hypothesis until now was "no".

If it's still "no", then the precompiled binaries will have to be
distributed uncompressed.

The question about the "make install" default still stands, though.

Re: Compressing the Lisp and Info files in the MS-Windows installation

[Stefan Monnier]

> The question about the "make install" default still stands, though.

I don't see any reason to behave differently under Windows than under
Posix platforms, in this respect.


        Stefan

Re: Compressing the Lisp and Info files in the MS-Windows installation

[Eli Zaretskii]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Cc: emacs-devel@gnu.org
> Date: Tue, 16 Apr 2013 09:10:17 -0400
> 
> > The question about the "make install" default still stands, though.
> 
> I don't see any reason to behave differently under Windows than under
> Posix platforms, in this respect.

Neither do I, FWIW.

Compressing the Lisp and Info files in the MS-Windows installation

[Stephen J. Turnbull]

Eli Zaretskii writes:

 > If the 'gzip' program is available, the Posix installation procedure
 > ("make install") by default compresses [stuff].  Are there any
 > reasons not to do the same when Emacs is installed on MS-Windows?

Wouldn't the sensible thing to do on Windows be to use zip, which in
some form seems to always be available?  While it doesn't directly
address Drew's gripe that it's a PITA for the user who wants to grep, it
does make it easier to undo the operation since there's only one zipfile
for the elisp and one for the info.

Many programs will now treat a zipfile as a filesystem mounted at the
zipfile's path, as well.  Dunno if grep or Emacs can do that, but
apparently it's a pretty standard thing (especially on Windows, where
Explorer has treated a zipfile as a directory rather than an archive
since Windows XP I think).

Re: Compressing the Lisp and Info files in the MS-Windows installation

[Eli Zaretskii]

> From: "Stephen J. Turnbull" <stephen@xemacs.org>
> Cc: emacs-devel@gnu.org
> Date: Tue, 16 Apr 2013 17:18:18 +0900
> 
> Wouldn't the sensible thing to do on Windows be to use zip

Maybe so, but it isn't trivial: Emacs doesn't support transparent
decompression of zip-compressed files as it does with gzip (and other
formats that compress single files).  So it's a separate project, and
not a trivial one, i.e. not limited to the configury stuff, which is
what I'm working on now.

> [zip] which in some form seems to always be available [on Windows]?

That's actually false, or at least not true in the sense you seem to
assume or imply.

What _is_ true is that Windows nowadays supports zip archives out of
the box in its File Manager (a.k.a. "Explorer").  But that support
comes from inside the application; there isn't any zip.exe or
unzip.exe when you install Windows, unless you install some optional
3rd-party package.

From my experience, that 3rd party packages tend to be GUI
applications, which Emacs cannot use anyway.  Not sure how this
compares with what others see.  (I personally use InfoZip's zip all
the time, but that's me; I see others like 7z etc.)

> While it doesn't directly address Drew's gripe that it's a PITA for
> the user who wants to grep, it does make it easier to undo the
> operation since there's only one zipfile for the elisp and one for
> the info.

Having the Info files in a single zip archive will need also support
from Texinfo's stand-alone Info reader, before it can be considered as
acceptable.  By contrast, gzip-compressed Info files are already
supported by Texinfo.

> Many programs will now treat a zipfile as a filesystem mounted at the
> zipfile's path, as well.

That's true, and so does Emacs.  But to grep this, we need _grep_ to
support that, which it doesn't.

> Dunno if grep or Emacs can do that

It cannot.

> on Windows, [...] Explorer has treated a zipfile as a directory
> rather than an archive since Windows XP I think

True.