[pracjourn-forum] For our error collection

[Peter Flynn]

On Fri, 2005-09-09 at 12:54, Peter Flom wrote:
> Peter Flynn wrote
> The error message should say something like
> 
> "You have used the math command \: in normal text, so I'm changing
>  into math mode"
> 
> But there is zero chance of the error messages ever changing...
> >>>
> 
> That would be a much better error message.  

Really, it's down to editor/processor interfaces to interpret TeX's
error messages and display something more reasonable.

Hmmm. A little script that could sit as an icon on your desktop
would do: just drag'n'drop your log file onto it and it would
display something sensible. I might toy with this in my copious
spare time during those long winter evenings I hear so much
about...:-)

> >From conversations at the PracTeX conference, I got the feeling that
> the TeX error messages are set in stone.  Is this true of LaTeX error
> messages as well?  

TeX messages have to stay as they are because they are among the output
checked by the TRIP test, which is used to certify versions of TeX. And
that was all defined by the author, and that isn't going to change 'cos
that is what defines a program which is allowed to call itself "TeX".

LaTeX messages can change, certainly. They did between LaTeX 2.09 and
LaTeX2e, and they may well do so again.

Unfortunately this particular one was a TeX message.

> I think this may be one of the big barriers to wider use; perhaps
> bigger than some long-time users realize.  People in math and computer
> science are used to a certain mode of thought  CS people are used to
> writing code and debugging it; math people are used to (I think) a
> psychologically similar process of trying one thing after another.  

Succinctly put.

> If people with this mode of thinking see a cryptic error, they may get
> annoyed, but also they are used to viewing this sort of problem as a
> normal process, and even view it as a challenge.  Social science people,
> and people in fields like education, do NOT have this mindset, and are
> likely to view such error messages as reasons to stop using LaTeX 

Yes. Many of them even panic. This is why the interface (usually the
editor) is so important, and most editor-writers don't realize this
yet. 

> Would the same "error messages set in stone" apply to commercial
> versions?

No, they are subject to Marketing, and don't have any predefined metric
for robustness as TeX does.  Marketing people don't give a tinker's spit
if the program falls flat on its face, so long as it looks pretty.

Somewhere between the two is probably about right.

> Or what about the messages that you get when you type H for additional
> help?

Those are designed for terminal console users. They still work, but
most users have never even seen a terminal window, let alone used one.
These messages too should be covered up by using /nonstopmode, and
left to the editor interface to interpret.

My €0.02

///Peter