Unfortunately Google has decided to remove the ability to view source from Google Groups, so we can't see the return-path for https://groups.google.com/g/comp.lang.perl/c/t4RumjajsXA/m/7..., one of the earliest messages he posted from netlabs.com, so we can't see what NNTP server he was using at the time. (I guess that's what we get for letting Google take the responsibility for "making information universally accessible": we have no recourse when they decide that means making previously public information inaccessible.)
The Wikipedia article says he released the first version of Perl when he was still at Unisys, citing the first edition of Programming Perl, which I don't have. The second edition (01996) is silent on the question. It also cites https://www.oreilly.com/pub/au/148, which does say that and presumably is at least subject to Larry's veto.
So, at any rate, JPL was funding Perl development from at least 01987 to 01991, four years, almost the same amount of time that Guido van Rossum was working at CNRI, 01995 to 02000. But you're probably right that Guido had to write grants and write progress reports on Python, and Larry didn't on Perl; he had the discretion to just do it. Also, I suspect this was true when Guido was at CWI too, as you say. AFAICT the only research paper Guido published at CNRI that was about Python was the CP4E paper.
I don't think it's accurate to say, "Perl is aimed much more at text processing and not linear algebra, while Python is more general purpose in this respect." PDL/perldl is from 01996, just the year after Numeric in 01995, and Perl 5 offers pretty much exactly the same set of facilities as Python for this sort of thing (dynamically loadable language extensions, operator overloading, dynamic typing --- though I guess at least Python's indexing syntax is more comfortable, because x[i:j, ..., 3] is a valid Python expression that preserves all that indexing structure, and has been since at least 1.5.2).
If memory serves, PDL was a lot better at 3-D plotting than Numeric was when I first tried it in about 01998; it could pop up a rotatable 3-D plot in an X window, and Numeric couldn't.
I think what happens is that a lot of people started working on Numeric (including those you mention --- although keep in mind dabeaz was also working on Perl's libraries and infrastructure!) and so it started getting better faster than PDL. Part of this was that Python is just a more pleasant, less clumsy language, so people chose it when Perl didn't have a killer advantage, like native mutable strings or a relevant CPAN module. But that's not about Python being well-suited for linear algebra; its only linear-algebra-specific feature is Python 3's @.
I think it would be very hard to find anyplace in 01995 or later that had Unix systems and didn't have huge piles of Perl. You're probably the person in the world most aware of the shortcomings of the primary alternative in the early 01990s (ksh/pdksh/bash). But it's also true that lots of Perl-heavy sysadmin shops never got into writing "real programs" in Perl like Slic3r, Perlbal, Movable Type, and Frozen Bubble, and so Perl didn't show up in their job descriptions. And nowadays most of those huge piles of Perl are sort of regrettable, and regretted.
Rich $alz reposted Perl in 01988, bearing a "1987" copyright date, but I think the first version really was released in 01986; at any rate by this point Larry was definitely at JPL: https://www.tuhs.org/Usenet/comp.sources.unix/1988-February/...
Unfortunately Google has decided to remove the ability to view source from Google Groups, so we can't see the return-path for https://groups.google.com/g/comp.lang.perl/c/t4RumjajsXA/m/7..., one of the earliest messages he posted from netlabs.com, so we can't see what NNTP server he was using at the time. (I guess that's what we get for letting Google take the responsibility for "making information universally accessible": we have no recourse when they decide that means making previously public information inaccessible.)
The Wikipedia article says he released the first version of Perl when he was still at Unisys, citing the first edition of Programming Perl, which I don't have. The second edition (01996) is silent on the question. It also cites https://www.oreilly.com/pub/au/148, which does say that and presumably is at least subject to Larry's veto.
So, at any rate, JPL was funding Perl development from at least 01987 to 01991, four years, almost the same amount of time that Guido van Rossum was working at CNRI, 01995 to 02000. But you're probably right that Guido had to write grants and write progress reports on Python, and Larry didn't on Perl; he had the discretion to just do it. Also, I suspect this was true when Guido was at CWI too, as you say. AFAICT the only research paper Guido published at CNRI that was about Python was the CP4E paper.
I don't think it's accurate to say, "Perl is aimed much more at text processing and not linear algebra, while Python is more general purpose in this respect." PDL/perldl is from 01996, just the year after Numeric in 01995, and Perl 5 offers pretty much exactly the same set of facilities as Python for this sort of thing (dynamically loadable language extensions, operator overloading, dynamic typing --- though I guess at least Python's indexing syntax is more comfortable, because x[i:j, ..., 3] is a valid Python expression that preserves all that indexing structure, and has been since at least 1.5.2).
If memory serves, PDL was a lot better at 3-D plotting than Numeric was when I first tried it in about 01998; it could pop up a rotatable 3-D plot in an X window, and Numeric couldn't.
I think what happens is that a lot of people started working on Numeric (including those you mention --- although keep in mind dabeaz was also working on Perl's libraries and infrastructure!) and so it started getting better faster than PDL. Part of this was that Python is just a more pleasant, less clumsy language, so people chose it when Perl didn't have a killer advantage, like native mutable strings or a relevant CPAN module. But that's not about Python being well-suited for linear algebra; its only linear-algebra-specific feature is Python 3's @.
I think it would be very hard to find anyplace in 01995 or later that had Unix systems and didn't have huge piles of Perl. You're probably the person in the world most aware of the shortcomings of the primary alternative in the early 01990s (ksh/pdksh/bash). But it's also true that lots of Perl-heavy sysadmin shops never got into writing "real programs" in Perl like Slic3r, Perlbal, Movable Type, and Frozen Bubble, and so Perl didn't show up in their job descriptions. And nowadays most of those huge piles of Perl are sort of regrettable, and regretted.