Message-Id: <199402041210.AA12123@mail.cs.tu-berlin.de>
Reply-To: Mailing list for the LaTeX3 project <LATEX-L@DHDURZ1.BITNET>
Date:         Fri, 4 Feb 1994 12:10:29 GMT
From: David Rhead <David_Rhead@VME.CCC.NOTTINGHAM.AC.UK>
Sender: Mailing list for the LaTeX3 project <LATEX-L@DHDURZ1.BITNET>
To: Multiple recipients of list LATEX-L <LATEX-L@DHDURZ1.BITNET>
Subject:      Additional features
Status: R

The "will it be backwards compatible?" dilemma that seems to be troubling
Mike Piff is not peculiar to LaTeX.

It seems, to me, to be pretty typical of the dilemma involved in
introducing any new software into user service:
*    a new Mark of the NAG Library.  If there is never a new Mark, people
     will never get improved routines.  If the new Mark has all the old
     routines, there will be a lot of "dead wood", and the manual will have
     to be even more enormous than it is now, to continue documenting the
     "dead wood".  But if the old routines are withdrawn, some old programs
     will stop working.
*    a new version of SPSS.  Up to a certain point, SPSS could be modified
     in line with its "punched card" ancestry.  To get an "interactive
     version", there had to be a major change which was not "backwards
     compatible".  Otherwise SPSS would have ceased to improve.
     In such circumstances, our computing centre might have both versions
     available to end-users in parallel (via different commands) and let
     the users migrate to the new one over 12 months.
*    What would happen now if you tried to compile a Fortran II program?  I
     haven't tried it, but I'd guess that a Fortran 90 compiler would fail
     to recognise certain Fortran II structures.  (And if the old program
     relied on a one-trip DO loop, it might produce different results
     under a newer compiler.)
In the NAG case, they keep obsolete routines in the Library for an overlap
period (one Mark = 18 months, I think).
In addition, any particular computing service that uses these products
probably has additional flexibility about when they introduce a new
version.  (In a university, "the summer vacation" is often a good time,
which cuts down the number of people who had to re-learn.)

So, there are various techniques (overlap period, delay in introduction)
for minimising the amount of re-writing (in these examples, of subroutine
calls, SPSS code, Fortran code), but ultimately the "least bad" thing is
that people who are "years behind" have to do some re-writing.
This, it seems to me, is fairly normal in the introduction of any new
software into user service.

Getting back to LaTeX3 ...
The question of
     If it is a choice between "getting things right" or "backwards
     compatibility", does one insist on "backwards compability"?
has been asked before, and I think the consensus was
     No.  Get it right.

Where does that leave anyone who wants
>    A reassurance!  \@startsection is cast in stone and will never
>    disappear or change its meaning again.
etc.?

Well, firstly, it leaves you in a better situation that with the commercial
products I've mentioned above.  LaTeX 2.09 is public domain, so you can
keep it on your computer for as long as you want.  If you want your
existing code to work forever, you can at worst keep it working by keeping
LaTeX 2.09 available.  (With the commercial products such as NAG, SPSS and
Fortran compilers, there would be licensing considerations.) Nobody is
going to turn up in Sheffield to destroy LaTeX 2.09 before you yourself are
ready to move off it.  Nobody is going to force you onto LaTeX2e or onto
LaTeX3.

However, if you have a "wish list" of improvements that you'd like, and the
idea at LaTeX3 is to give priority to "get it right" over "backwards
compatibility", you may not get the improvements without sacrificing some
"backwards compatibility".  (E.g., LaTeX 2.09 has "same vertical space
before and after lists", which makes it difficult to distinguish paragraphs
when block paragraphs are used.  I put extra \vspace in my .tex files when
necessary with LaTeX 2.09.  If LaTeX3 "gets it right" I'll have to take the
\vspace out of my .tex files.  But I'd regard it as worth it for the
overall improvement.) You'd be in an analogous position to someone who has
a Fortran II program with a one-trip DO loop and who wants the features
available in Fortran 90.

If the worst comes to the worst, you could have 2 versions of LaTeX
available in parallel so that you don't place your existing investment in
specialized .sty files for exam papers etc.  at risk.  Perhaps aim to check
out your .sty files with LaTeX2e over the next 18 months, and then aim to
update things for LaTeX3 in the 18 months after that appears.  It's
analogous to what you'd have to do if you had a Fortran program calling a
particular NAG routine, and NAG gave notice that that particular routine
was to be superseded by something better.

                                                                David Rhead