[This thread just started on e-lang, but I thought I'd cross post it to cap-talk. Ping, I cc'ed you in case you're not on cap-talk. --MarkM]
At 05:10 PM 5/6/00 , Ka-Ping Yee wrote:
>Have you all seen this paper?
>
> http://cm.bell-labs.com/who/rob/utah2000.ps
>
>It seems worth reading and discussing here, and perhaps on the EROS list.
Excellent paper, until he gets to "Things To Do" / "Things To Build". Despite his attempt at pessimism, here he asks people to just go back and do the kind of wonderful innovative systems research/building they did in the old days, with the implication that it won't be irrelevant this time. This after explaining well why todays world has tended to make such work irrelevant, but not saying what difference he proposes.
And yes, EROS and E are both examples of the kind of work he calls for. However, though he correctly predicts that in both cases most of the work will need to be on connecting these to existing bad standards out of our control (at least Java & Unix respectively), he misses that this is where the *hope* is: That, because of Turing universality and other tricks, a system may both (seem to) be made familiar enough to have a hope of surviving & prospering, and at the same time, inside, be pure clean and innovative.
To make this case, one has only to look at his comparison of the 1990's vs 2000's high end workstation (on page 4). The software innovations are labelled "Netscape", "Java", and "Perl (a little)". Taking the first two (Someone else take Perl. Please.), and taking "Netscape" to stand for the Web, how did these two innovations succeed despite all of Pike's correct explanation for why todays world squashes software innovation? Because they were much more innovative than they seemed. The early web seemed to be just "ftp readme files you could click on", and Java succeeded because it seemed like little more than a better C++.
This hope is Subversive Innovation. Build something ordinary seeming enough to be adopted & used, but extraordinary enough to really change things once it has been.
Btw, the software area where there is true systems innovation, building, and research in the old style is cryptography. Although X509 may be ending that era in the same way Unix ended the earlier one.
Cheers,
--MarkM