Skip to content

maketitle.pl for Perl 4.004

7 messages · Peter Kleiweg, Kurt Hornik, Brian Ripley +1 more

Original
Peter Kleiweg
# 
Kurt Hornik scribeva...
I am sorry to hear that.
Unfortunately, that is not an option at the moment. I guess most
people working with R are professionals with the means to work
with relative new hardware. I have to make do with a very old
system, a disk drive that is almost full, so no possibility to
compile a recent version of gcc, or upgrade anything that
demands a C compiler that is not from last century.

I write quite some software myself. Working with an old system
like I have has the advantage that I never hear people complain
that my software doesn't work on their system because there
system is too old.

But I have to accept that my system, for the time it keeps
working, is becoming state-of-the-art-long-gone.
Kurt Hornik
# 
Following the same reasoning, wouldn't one always stay with R 0.14 or
similar?  Note that e.g. requiring Perl 5.005 only has implications for
*developers* and not for users because we have no dependency on Perl at
run-time.  And the hope would be that developers have access to software
which is 4 years old and typically available in pre-packaged binaries.

-k
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
Brian Ripley
#
On Wed, 1 May 2002, Peter Kleiweg wrote:

            

      
      
      
    

        
I think we are almost all amateurs at computing.  Certainly most of my R
development is done on machines bought with my own money, the main ones
being 4 and 3 years' old respectively.

I think the solution may be to not upgrade R either!

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Peter Kleiweg
    

    

      
      #
    

  


  

      
Kurt Hornik scribeva...

            

      
      
      
    

        
Of course R should be developed further. But requiring Perl
5.005 does effect end users. I have stopped downloading
binaries long ago, because they never work on older systems. So
just to use R, I have to compile R myself. And just installing
the recommended packages for R 1.5.0 needs Perl 5.005.

And what about new or upgraded packages? Do you ask users
to reinstall R just because they want to add a package, or do
you keep packages compatible with older versions of R? I always
go for backward compatibility as much as possible when I'm
developing software.

Of course, my system is very old. There won't be many people
left working with such an old system. I can understand you have
to draw the line somewhere. But the package 'tcltk' won't work
on a system just over two years old. I guess this will be a
problem for many people.

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      A.J. Rossini
    

    

      
      #
    

  


  

          

      
      
      
    

        
peter> Of course, my system is very old. There won't be many people
    peter> left working with such an old system. I can understand you have
    peter> to draw the line somewhere. But the package 'tcltk' won't work
    peter> on a system just over two years old. I guess this will be a
    peter> problem for many people.

Sure, but there are perfectly great stat language/packages which are
small, compilable, and work like a charm, without some of the great
benefits of R.

XLispStat comes to mind, and in fact, has had functional
cross-platform widgets for a while.

Of course, there is always S-PLUS.  It doesn't require building,
and may be comparable to a full installation of R, depending on what
that means to you.

It seems like it might be worth reconsidering your software platform
in light of your hardware requirements.

best,
-tony

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Peter Kleiweg
    

    

      
      #
    

  


  

      
A.J. Rossini scribeva...

            

      
      
      
    

        
It seems some people think I am criticising the development of R
because it won't work (in some time) on my machine. That is not
the case. My comments are meant in general, as something to
think about. My only interest is in helping the development of
R, with the limited suggestions I can offer.

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      A.J. Rossini
    

    

      
      #
    

  


  

          

      
      
      
    

        
>> It seems like it might be worth reconsidering your software platform
    >> in light of your hardware requirements.

    peter> It seems some people think I am criticising the development of R
    peter> because it won't work (in some time) on my machine. That is not
    peter> the case. My comments are meant in general, as something to
    peter> think about. My only interest is in helping the development of
    peter> R, with the limited suggestions I can offer.

No, the general issue is that software construction is difficult.
There are tradeoffs, and your comments actually are more specific than
general, bringing up a situation which is not necessarily common.

People do want to have ancient software supported.  I.e. with ESS, we
ended up supporting truly ancient, unbuildable (on modern compiler
platforms) software, i.e. Emacs 19.28, because it was what DEC
shipped, and a vocal (SMALL) minority wanted it.  It crippled (but
didn't halt) development of some rather useful features, because of
lisp-level functionality that was introduced by the end of 19.xx
series, Emacs 19.34, which was deployed around 1995 or so (I'd have to
check dates, but its ancient).  In retrospect, we should've nuked
support in 1999 (not 2001).

I don't feel that your comments are critical, but I do feel that they
aren't addressing the point (well, it's one, anyway), which is how to
get a usable system built and deployed given the constraints of all
(developers and users) involved.  Your constraints are not really
solvable (want R 1.5+, and Perl 4..), and hence, I suggested (modern,
functional) tools which may solve it, of course introducing a
completely different (and unevaluated) set of tradeoffs...

best,
-tony