search.cpan.org

» see search.cpan.org as it lived, on the wayback machine

~ the obit ~

search.cpan.org went live in early 1999, when I was fifteen and the modem at home was a 33.6k. Perl was a word I knew from the CGI scripts that made guestbooks work, not a language I had ever had a module problem in. What follows is an obituary for the most boring and most necessary page in somebody else's working day, nineteen years of it.

CPAN itself is the archive: the Comprehensive Perl Archive Network, a set of mirrors holding every published Perl module, and one of the first package ecosystems any language had. An archive of files is not usable on its own, though. What Graham Barr built was the front door, the thing that let you type a module name into a box and get back its documentation, its version history, its dependencies and the author who maintained it. For most of two decades, that host was how Perl was actually used.

The hosting history reads like a map of how the open-source web was held up in those years. It started on a Solaris machine at Washington University in St. Louis. American mirrors came from YellowBot and from Phyber, and a European mirror was run by Digital Craftsmen for roughly the last decade of its life. Operations eventually moved onto the Perl NOC, the volunteer infrastructure team that keeps the language's own services running.

~ the rap sheet ~

~ nineteen years of accumulated perl ~

The retirement announcement in May 2018 is candid in a way that corporate shutdown notices never are. Maintenance had become a burden, it said, and most of the site was running 2005-era Perl code. Nobody had broken it. Nobody had abandoned it either. It was simply a working system written against the assumptions of a different decade, maintained by volunteers, still serving a load that mattered.

MetaCPAN had been the real alternative for years by then, with a modern search index and a team around it. The announcement's actual claim was narrower than obsolescence: there were now two front doors to the same archive, and only one of them had anyone left to look after it.

~ the switch ~

The Internet Archive brackets this to a single day. At 04:19 UTC on 24 June 2018 the host served its own homepage one last time, 2,062 bytes of it. By 19:03 UTC the following evening the same URL was a 301 of 528 bytes. The Perl NOC's follow-up post is dated 26 June and opens with "Yesterday, we threw the switch."

What the MetaCPAN team did with the month between announcement and switch is the part worth recording. They mapped the old URL structure onto the new one so that nineteen years of links in mailing list archives, Stack Overflow answers and README files would keep landing on the right page. Olaf Alders, who founded MetaCPAN, wrote the migration up on perl.com.

~ a host that is now a direction ~

Ask for search.cpan.org today and you arrive at metacpan.org, as you have for eight years and four months. The old service is gone in the sense that none of its code runs and none of its pages exist. The name still has a job: it points the links made while it was alive at the thing that replaced it. That is the kindest ending available to a website.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »