I'm not on the PMdev team, so these are only my own observations, and in no way authoritative:
Nodelets seem to only manifest themselves when there's relevent content. For example, the "Functions" nodelet does not appear if there's no perl code in the node, or if you're at the root of a section. So the CPAN nodelet, for example, hasn't been present even though selected by you, I, and other monks presumably because of the recent background problems within CPAN peering(?).
Something I forgot to include in the root of this thread is that I finally got around to eliminating my pet theory* of problem within one or more specific internet distribution points, by pinging** the PM host at the same time as I was experiencing page load delays. The ping responses were consistantly good (good for dialup, at least) at the same time as PM response varied from horrible to fine. So when tye mentioned nodelets, the light bulb clicked on! 8^)
cheers,
Don
striving toward Perl Adept
(it's pronounced "why-bick")
* Fastolfe independantly arrived at the same theory here
** I prefer to know ping response times _before_ doing a traceroute. If the end-to-end ping response is fine, then I don't care about intermediate responses, since my real-world application _isn't_ getting responses from intermediate routers.
Also, for more meaningful ping info, specify a near-mtu ICMP packet size. With Linux, I did "ping -s 1000 www.perlmonks.org", on win32 the switch would be "-l" (ell, not one). The default ICMP echo-request packet is 56 bytes, whereas almost every TCP packet carrying HTTP payload will be MTU. On my dialup connection it made the difference between ~200ms and ~400ms average response. | [reply] |