Debian Policy based on the wrong technical assumptions
I fear the Documentation Policy is being based on the wrong technical
assumptions.
The fact is that what slows down HTML in an old system is not a web server or
the cgi converters. It is the browser!
Please take a moment and read the following timings made in a 386 SX-25 with
5MB of RAM. It will help you understand what is at stake. I did the test as
fairly as I could and I claim to be a honest person.
/tmp$ time lynx -dump \
http://localhost/cgi-bin/man2html/usr/man/man1/telnet.1.gz >/dev/null
5.98user 82.76system 1:36.40elapsed 92%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (360major+74minor)pagefaults 0swaps
/tmp$ time man -l /usr/man/man1/telnet.1.gz >/dev/null
Reformatting telnet.1.gz, please wait...
101.25user 24.03system 2:11.48elapsed 95%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (1795major+513minor)pagefaults 125swaps
As you can see, man was more CPU intensive and took more time than lynx with
on-the-fly conversion.
/tmp$ time snarf \
http://localhost/cgi-bin/man2html/usr/man/man1/telnet.1.gz - >/dev/null
0.12user 0.12system 0:03.30elapsed 7%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (74major+16minor)pagefaults 0swaps
Snarf was useful to show that the overhead of boa in a 386 SX-25 is _minimal_
compared both to man and lynx.
/tmp$ time lynx -dump telnet.html.gz >/dev/null
6.81user 83.47system 1:33.22elapsed 96%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (550major+147minor)pagefaults 0swaps
Lynx accessing the pre-formatted file was only 3 seconds faster!!
/tmp$ time lynx -dump telnet.html >/dev/null
6.16user 81.58system 1:30.76elapsed 96%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (326major+68minor)pagefaults 0swaps
Even more. Lynx accessing a pre-formatted and uncompressed file was only
6 seconds faster than on-the-fly conversion!!!
If we keep the preferred online format as HTML we either have to live with
the slowness of web browsers or write a simpler and faster one. But let's
not try to be deceiving here. On-the-fly conversion adds very little
overhead and in fact improves the speed compared to man+groff.
Please base policy on facts, not on pre-judgements.
Thanks,
Fernando.
--
TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to
debian-devel-request@lists.debian.org .
Trouble? e-mail to templin@bucknell.edu .
Reply to: