X-VM-v5-Data: ([nil nil nil nil nil nil nil t nil]
	[nil nil nil nil nil nil nil nil nil nil nil nil nil nil nil nil])
Date:         Mon, 09 Jul 90 13:51:09 CET
From: Rainer Schoepf <BK4@DHDURZ1.bitnet>
Organization: Inst. f. Theor. Physik d. Univ. Heidelberg
Subject:      Re: Permanence of .aux files
To: LaTeX-L Mailing list <LATEX-L@DHDURZ1.uucp>
Status: R
X-Status: 
X-Keywords:                  
X-UID: 172

As Barbara points out, file systems with version numbers are helpful,
but not a full solution. Murphy always wins...yes, but I prefer to call
it the law of conservation of trouble.

I can see three reasons for .aux files beingincompletely written:
- bad information passed from within LaTeX, such as the long caption
  problem: here we can avoid this by tricky TeX programming (should
  we really do this?)
- interruption of writing, e.g., by an overflow of TeX's memory or
  by typing `x' while in a break loop: this can be avoided by the
  two step approach proposed by Nelson (unless the interruption occurs
  while the second file is written which is highly unlikely).
  Actually, this is the reason why .toc are nearly never damaged.
  (They may be empty, but this can be avoided.)
  One could do better if TeX had a system interface.
- external causes, such as running out of disk space: no chance.

The two step (auy-->aux) approach has another advantage: it is possible
to write information (namely to .auy) while LaTeX is still reading
(from .aux).  This can be used to read only the information into memory
that is really needed.

-Rainer