65K was the prevailing annual limit, though the government was apparently lax about staying under it. The H-1B was temporarily increased to 195K, the last time in 2000 October, well after the economic depression had begun (Y2K bust in 2000 January, dot-com shake-out from 1998 through 2002, Nasdaq crash 2000-03-10, up-tick in lay-offs in 2000 July & August). It was quickly followed by massive tech lay-offs in 2000 December.
In addition to the 65K normal limit, and the 20K limit for those with advanced degrees earned in the USA, there are an additional 10,500 E-3 visas for similar kinds of work that are set aside for people from Australia, and none of these limits apply to employment by government, colleges, universities, or non-profit organizations.
While the tech executives continue to disingenuously argue that H-1Bs are only to bring in people with rare skills ("super-stars", "the best and the brightest", "pre-eminent"), in actuality H-1Bs were used to flood the US tech job market with mediocre but cheap talent at the margins in order to drive down compensation and employment of the many perfectly capable Americans already in the field. Of course, if they were only bringing in super-stars, the limit would need to be on the order of a mere 1K to 6K; there just are not that many super-stars in the world.
If there really were a shortage of science and tech talent, economists would expect to see significant increases in total compensation, above inflation and above other professions. That did not happen.
We'd expect to see more, and more generous, relocation offers. Instead, we see fewer.
We'd expect to see more, and more substantial, new-hire training and education offers. Instead, we see fewer.
We'd expect to see barriers to employment melt away. Instead, since the 1980s we have seen increasing barriers (various tests, series of telephone "interviews" followed by genuine interviews, stress interviews...).
If there were a "shortage" we'd expect to see human contact names, human voice telephone numbers, e-mail addresses read by humans, fax numbers, and street addresses in nearly all ads, as well as web addresses. Instead, we see fewer.
We'd expect employers to more conscientiously examine resumes and applications. Instead we see a decreasing tendency (and ability?) to do so, as they play a mass numbers game.
We'd expect tech firms to expand college recruiting to more of the thousands of colleges and universities in the USA. Instead, they limited themselves to fewer than 30 (Gates recently mentioned that M$ actively recruits from a mere 26).
We'd expect to see a higher fraction of new grads in these fields to be snapped up by employers. Instead, we saw some 35% passed over.
We'd expect to hear recruiters complain of the small numbers of applicants. Instead, we saw them complaining of receiving hundreds of applicants for each position.
Even during the so-called "boom", we were reading reports of unemployment rates into well into the double digits among older programmers, whether or not they had conscientiously engaged in "continuous learning". In the late 1990s it was reported that going to job fairs, where recruiters would become aware of the ages of applicants early in the process, could add months to a job search.
Students have shied away from the field because of the drop in job security and expected real (inflation adjusted) life-time earnings which has worsened since the 1980s but really took a dive during this depression.
Go to the job sites (Monster, Dice, computerwork, etc.) and you'll see that there are many job ads for body shops and few for full-time permanent hiring by end employers (sometimes called direct-hires). It is also interesting that none of these sites offer a simple and direct way to select ads for full-time permanent employment only or vice versa (DICE comes closest). This is not a sign of a healthy science and tech job market.