Re: A stats library (announcement)

Reef Fish wrote:
> Russell.Martin@xxxxxxx wrote:
> > Mr. BK of Baltimore, MD wrote:
> > >
> > > Allow me to rant some more about stats packages and why learning a
> > > dozen of them drove me to waste time writing a general-purpose stats
> > > library.
>
>
> > Your points (or rant :-) ) are well taken, but I think Babel has
> > broken upon us and there will be no going back,
>
> True to a large extent for many reasons you've advanced (all they
> have at work; wheel-re-inventers; limited exposures to what had
> already been built, etc.) but your stereotypical characterization
> is both myopic and unnecessary.
>
>
> > they know, people who can manage through contortions to do what they
> > need with the software they know (you know the type: "I can write
> > this text editor in FORTRAN :-) )
>
> FORTRAN was a wunderkind after I had to learn machine language,
> Symbolic language, Assembler, and other primitive tools, but
> FORTAN has been OBSOLETE to any one in the know about languages
> for almost half a CENTURY!

Of course, FORTRAN aficionados would say your knowledge of the
language is dated and that FORTRAN has been modernized, expanded,
and improved. You should argue that with them, since I seldom
use it any more (although I am about to have to use it on a
project I'm "inheriting"). My point was about the people who
write software in the language they know no matter what, so you
can substitute writing a general circulation model in Korn
Shell or whatever example you prefer.

>
> > Even a good design (your library or Speakeasy
> > or whatever else someone wants to proclaim as wonderful) would have
> > to be very inclusive (small work function for many people)
>
> Your generalities do not apply to SPEAKEASY.

The generalization that it has a work function to overcome and
because of that many people would not move to it does apply.
And because of human nature, the work function doesn't have
to be large. In fact, all it has to be is perceived of as
significant (which may be more the fault of the environment
that the software itself), and clearly it is difficult to
overcome such externalities. That is not necessarily a
reflection on the language, but for whatever reason Speakeasy
has not conquered the computing world. The problem may have
be the promoters possibly thought it was so good that it
promoted itself. I don't know. The landscape is littered
with software that was good in its time and was supplanted
by, IMO, inferior software. Often it was, broadly speaking,
marketing as much as anything that made the difference, as
far as I have seen. All that goes to show that factors in
addition to how wonderful the language is also play a part
in success or failure, and Speakeasy is not immune.

> It invented many
> NEW wheels in the 1960s, and most of them are STILL NEW today, to
> all the other software mentioned by you or BK. It invented ONLY
> new and useful wheels, not re-inventing old wheels if others
> have already invented them and done them quite well.
>
> It makes USE of the some of the best features of the best libraries
> around already -- LINPACK, ICEPACK, NAG, and other well-known and
> well establish routines invented by others.
>
> I am more multi-lingual in somputing software than most computer
> jocks and hacks, though I don't consider myself one. I was VERY
> familiar with at least a dozen or two of the major statistical
> packages; very familiar with a dozen different LANGUAGES. It's
> only through such knowledge that I was perfectly satisfied NOT
> to seek or invent any new wheel since about 30 years ago, because
> there HASN'T been anything that came close to SPEAKEASY as a
> language, let alone better, except for certain very specialized
> tasks.
>
> In that respect, I offer myself as a counterexample to everything
> you've said in your post. :-)
>
>
> > By now I've probably have ticked off pretty much
> > everyone who holds an opinion about software, either explicitly or
> > implicitly by not mentioning their favorite, so I'll wish you good
> > luck with your efforts.
>
> You mentioned my favorite, but you also showed that you are almost
> ignorant of both the OUTnard and INards of that software. :-)

It brings back memories. I used it at Argonne in 1974 to obtain
some results that went in the first paper I ever coauthored. I
liked it a lot, but of course it was light years ahead of
anything I'd used up to that time. It was also a resource hog
(in terms of the total system I needed to use it for my project,
which may not have been completely the language's fault) and I
got taken to task by the assistant division director for spending
as much money as I did using it. I told him my boss had approved.

>
> It has many NEW WHEELS (that were left over from 30 years ago) that
> haven't been discovered by all the present and past generations of
> programmers and software hackers yet, while it has also taken
> advantage (rather than re-invent) may old and working wheels, in
> linear algebra, optimization techiniques, numerical analysis
> methods, etc.
>
> That's why I hadn't have to RELY on any statistical package or
> other software languages that I knew, and knew VERY well, for
> DECADES.
>
> So, there!
>
> Disclaimer: I hold no stock or sell the SPEAKEASY software, nor
> will I benefit directly or indirectly from anyone else purchasing
> or using that package. Just giving you my EXPERT opinion about
> computer languages and packages that are good for statistical
> computing.
>
> If you still don't look carefully or try it, it'll be your loss,
> but you got what you paid for in this newsgroup. :-)
>
> -- Bob.

Cheers,
Russell

.

Relevant Pages

  • Re: A stats library (announcement)
    ... >> FORTRAN was a wunderkind after I had to learn machine language, ... The reason that SPEAKEASY never caught on like much, ... >> NEW wheels in the 1960s, and most of them are STILL NEW today, to ... >> packages; very familiar with a dozen different LANGUAGES. ...
    (sci.stat.math)
  • Controlled Dynamism
    ... In reading the recent threads about the direction of scheme ... desires about what the language should be. ... additional "semantics" packages implementing other required ...
    (comp.lang.scheme)
  • Re: "R" Program - Executable?
    ... newsgroup for the "R" statistical programming language ... R is becoming an important language to learn for statistical computing ... The linear models packages are almost as powerful as SAS's ... FDA has strict rules about the validation of software. ...
    (sci.stat.math)
  • Re: A stats library (announcement)
    ... >> Reef Fish wrote: ... >>> FORTRAN was a wunderkind after I had to learn machine language, ... >>> Your generalities do not apply to SPEAKEASY. ... >> And because of human nature, the work function doesn't have ...
    (sci.stat.math)
  • Re: fc5: install everything?
    ... I counted about 800 clicks to get a complete install by clicking on all ... of the separate packages, (not counting the couple of reinstalls because ... Thereafter having to update umpteen unused OOo language modules ... one for all apps and a separate one for all languages. ...
    (Fedora)