Re: [Groff] Compression support for files?

[Steffen Nurpmeso]

Hello Ingo,

Ingo Schwarze <address@hidden> wrote:
 |Also, to make life easier for yourself, never put two or more
 |logically unrelated changes into the same diff, in particular

well.. it's not really the same diff though i also have an alias

  alias.mail-patcho !git format-patch --output-directory ~/tmp/git-patcho

(And i also have

  [sendemail]
  smtpserver = /Users/steffen/usr/bin/s-nail
  smtpserveroption = -t
  smtpserveroption = -Snodot
  smtpserveroption = -Asn_y
  assume8bitEncoding = UTF-8

but unfortunately my shitty mua cannot yet swallow the necessary
Message-Id: etc. fields but throws them away, for quite some time.)

 |not into large ones (like this one).  The only exception are maybe
 |small diffs collecting several minor changes as long as all of
 |them are certainly trivial and uncontroversial.
 |
 |Mixing changes makes a diff harder to review for others,
 |harder to get feedback for you, and if you do get feedback,
 |it is likely do be unrelated and the discussion is likely
 |to get derailed, as you see here...

I'm sorry for the inconvience!  It was thought like this (in S-nail(1))

  ?0[ +track]? f
  >N 111 Steffen Nurpmeso  18:53 07-16 2181/ 65381 [Groff] Compression supp
  ?0[ +track]? !git co -b zzz origin/master
  Branch zzz set up to track remote branch master from origin.
  Switched to a new branch 'zzz'
  !
  ?0[ +track]? w  
  Enter filename for part 2 (text/x-diff): |git am  
  Applying: contrib/gideal/Makefile.sub: fix Bernd Warken (Public Domain)
  Applying: src/include/lib.h: add some infrastructure (Public Domain)
  Applying: Add class file_case to encapsulate FILE* (Public Domain)
  Applying: Allow compression for *all* files (Public Domain)..
  "/dev/null" [Appended] 26/978
  ?0[ +track]? !git lo origin/master..
  * f12789b ([.]) Allow compression for *all* files (Public Domain)..
  * 691552e Add class file_case to encapsulate FILE* (Public Domain)
  * 59e292a src/include/lib.h: add some infrastructure (Public Domain)
  * 3358bc9 contrib/gideal/Makefile.sub: fix Bernd Warken (Public Domain)
  !
  ?0[ +track]? 

and then giving you all possible opportunities..

 |I didn't come round to review your diff yet, but here are two
 |basic observations:
 |
 | 1) Disk space is rarely expensive nowadays,
 |    so i rather strongly dislike installing *anything*
 |    compressed.  Often it hinders searching with grep(1).
 |    Sometimes it hinders viewing files with cat(1).
 |    Often it slows down operation - decompression has to be
 |    done again and again, uselessly.
 |
 | 2) Even if such an option were disabled by default,
 |    i would rather not have the option than have it.
 |    No matter how you do it, it is likely to add some
 |    complication and another layer of indirection to the code.
 |    So i'd like to see a rationale which real-world needs
 |    this is going to help and why these needs cannot be
 |    satisfied in other, more trivial and UNIXy ways,
 |    like pipes involving the compression tools.
 |    Saving disk space is not a good reason, IMHO.

The original intent of Matthew Dillon i think was to be able to
blindly use the `.so' request and make it work regardless of
wether whatever file is compressed or not.  This is currently
absolutely impossible which is something that i stumbled over
myself too in the past (with the solution being not to use the
inclusion requests at all).

And i personally totally disagree with the disk space argument.
It may be cheap, but that doesn't mean i have it, nor that an
african child that gets sponsored a super-cheap netbook has, or
something.  My personal backup USB stick is 8 GB, and it is
sufficient to store a monthly backup plus the growing daily backup
until the next monthly backup.  This covers all my arena with
documents, own and foreign code, and the monthly backup even
includes the git(1) / git<-hg(1) full-history repositories of four
operating systems plus a full-history package manager history of
a Linux distribution.

Moreover, some files you simply don't need that often; a Unicode
font in the format that groff(1) understands can well be over 40
megabytes, which can be boiled down to less than the half of that.

Ciao Ingo,

--steffen