In 1989, there was a matrix model (subfield of string theory) conference at Rutgers, and Paul Ginsparg gave me the tex files for several preprints associated with the meeting. I
distributed these papers to friends (via email) who I thought might be interested.
Colleagues from other institutions then started to ask me for papers on string
theory more generally, by
authors at the IAS (my then home institution). I'd ask the authors for their papers, send them to the people who had asked, and then send them as well to people who had asked for
other papers previously, or whom I thought might be interested.
Slightly later on, I learned how to use an email exploder. I began to systematically expand the number of names on the list I was sending papers to. I also expanded the role of the mailing list from just a list which received papers I had, to a group of people who both received and contributed papers. In this way, it became a way for people to exchange papers more generally. I started asking people I knew to send me their papers. I also asked people I didn't know, if I saw their papers and their emails were available, for their papers (and simultaneously invited them to join the list). People also requested to join the list who heard of it via word of mouth; I would add them and also request them to send their papers. Often papers would go to one person who would print it out for the group (although some research groups requested me to send papers to all members individually, which I did). Eventually (by 1991 summer) it was reaching several countries and institutions (I do not have the list anymore, but have asked Paul to send it to me, as he does have it for the reason given below). I believe it had about 180 people (a small number by today's standards!) and reached over 20 countries.
I think part of the reason people were interested in joining is because I had papers they were eager to get (some people were still getting articles only when they came out in print, by surface mail). Also at one point two people put out papers on the list on the same topic within a few days of each other; my impression was that because of the worldwide reach of my distribution list people realized it could be a way to establish precedence for research results.
I was sending out the papers in emails, by hand. In the summer of 1991, Paul Ginsparg asked me why I didn't automate my system. We were at an Aspen Physics workshop (he explained later that he had been thinking about such a thing for a while at that point). I said that I did not know how to do so but that he was my guest to try. The next day he told me he'd written some scripts to do it (my understanding was that he had done so between our two conversations). We then made a plan to transfer my network to his new system later in the summer. While he was travelling he also began to announce that his automatic network was going to begin soon, and added more names this way. The automated list began on Aug. 14, 1991. I sent an email asking people on my list to switch over to his automated list (one person wanted the old mailing, which I did) and then stopped my mailings.