Path: utzoo!dciem!nrcaer!scs!spl1!laidbak!att!pacbell!ames!ll-xn!mit-eddie!genrad!panda!teddy!jpn
From: jpn@teddy.UUCP (John P. Nelson)
Newsgroups: comp.sources.d
Subject: Re: Poll on shar formats
Message-ID: <4838@teddy.UUCP>
Date: 3 Jun 88 14:19:10 GMT
Article-I.D.: teddy.4838
References: <868@fig.bbn.com> <1494@microsoft.UUCP>
Reply-To: jpn@teddy.UUCP (John P. Nelson)
Organization: GenRad, Inc., Concord, Mass.
Lines: 31

[Discussion of shar prefixing algorithms]

Back when I was mod.sources moderator, I modified my version of "shar"
to pre-scan each file for "dangerous" character sequences.  If no
"dangerous" sequences were found, then no prefixes were used at all,
and "cat" was used to extract the files, not "sed".  This both runs
faster, and makes it easier to extract the file when a "dumb" editor is
the only available means of extracting the files.

In a year and a half of moderating, not one file actually needed
prefixing.  Of course, I didn't repack any shars that I didn't need to,
but I ended up repacking about 1/4 of the submissions.  I did not get
ANY complaints from people about strange sequences causing corruption
of shars.

For some reason, people assume that a dot beginning a line is a
"dangerous sequence":  It is NOT!  What they are thinking of is a dot
ALONE on a line:  This causes some mailers to terminate reading the
mail.  It is silly to prefix EVERY line starting with a dot (nroff
source) because of this.

Other dangerous sequences are a line starting with "From", or a line
starting with the here-document end-of-file marker.  There may be
others, but I cannot recall any.  Note that a leading 'X' is not 
dangerous unless you are already using sed 's/^X//' to extract.

-- 
     john nelson

UUCP:	{decvax,mit-eddie}!genrad!teddy!jpn
smail:	jpn@genrad.com