MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
References: <19109.1284579373@cl.cam.ac.uk>
            <4C9122A1.4060008@morningstar2.co.uk>
            <4C91239B.2040301@morningstar2.co.uk>
            <OF4383141C.F0E37102-ON8025779F.006F4D59-8025779F.006F4D5C@mcs-notes1.open.ac.uk>
            <4C912F95.70503@morningstar2.co.uk>
            <OFF0867069.119AEB13-ON8025779F.0076844B-8025779F.00768450@mcs-notes1.open.ac.uk>
Message-ID:  <19601.55875.121644.485219@morse.mittelbach-online.de>
Date:         Thu, 16 Sep 2010 10:50:11 +0200
Reply-To: Mailing list for the LaTeX3 project
              <LATEX-L@listserv.uni-heidelberg.de>
Sender: Mailing list for the LaTeX3 project <LATEX-L@listserv.uni-heidelberg.de>
From: Frank Mittelbach <frank.mittelbach@LATEX-PROJECT.ORG>
Subject: Re: e-TeX low-level expertise needed Re: Allocation of registers
To: LATEX-L@listserv.uni-heidelberg.de
In-Reply-To:  <OFF0867069.119AEB13-ON8025779F.0076844B-8025779F.00768450@mcs-notes1.open.ac.uk>
Precedence: list
Status: R

Chris,

 > >>
 > I did a few tests. As the e-TeX manual says, you can't \insert beyond 
 > 254. \insert255 gives the special 'You can't \insert255' error (as 
 > without e-TeX), and anything higher is a bad register code. (The 
 > {<stuff>} seems to vanish.)
 > >>
 > 
 > Ah, but is that vanish b: or vanich c: :-) ??   
 > I think we should be told!  I suspect b: (ie the tokens in <stuff> are
 > never processed at all) but what a waste of tokens...!!  And the error
 > message or something should point this out to B.L.Euser 

you shouldn't be too hard on Peter (who I guess wrote that code) --
if eTeX replies with a "bad register code" error then I think that is good
enough even if perhaps not the best error message you can get. And frankly
what happens after a severe error is kind arbitrary so that it really doesn't
matter much in my opinion whether you will be told in the manual exactly how
the program recovers.

So in my opinion, if eTeX manual states that \insert only works on 0-254 and
then also goes and generates an error if you try differently then it does all
that is needed.

A consequence for us seems to me that indeed we can't do much more than
reserve a range for use with inserts at least while we live in coexistance
with code that uses inserts.  The boundary is reserving all lower registers,
but that is most certainly overkill and as the higher registers are supposed
to be less efficient we should probably not waste too much of the main
registers.

For now I think providing an additional 30 inserts is the right level

frank