don@clunk.UUCP (Don McKillican) (01/21/87)
I should like to thank the few people I have heard from so far with corrections to the map I posted: rayan@utegc for the UofT stuff, and emc@unicus and david@geac for corrections. The changes will be incorporated into the next version. Clunk had a uucp seizure this past weekend, so if you've sent me something, I probably didn't receive it: please try again. Now to my critics :-). In article <1504@lsuc.UUCP>, dave@lsuc.UUCP (David Sherman) writes: > Hmm. Thanks, Don. I didn't even suggest anyone try doing > a UUCP map, because (1) pathalias can give routings and (2) > the map would be so complex. Your attempt is nice, but is > (just for example) missing lsuc's connections to parkridge, > hcr, nrcaer, toram, maccs and sickkids, First of all, both your points are perfectly correct. But I think there is room for another point of view, too. Pathalias is okay, but there are a lot of sites which don't have it, don't use it, don't keep it up to date or any combination of the above. And there are even more users who don't know how to use it. And I personally happen to like a pictorial approach better. Besides, if pathalias tries to send me through lsuc, and I happen to know that lsuc is (perish the thought!) down, then I can look on the map for an alternate path. Pathalias isn't good at that. As to the sites you name, I actually had a reason for all of them. You will recall I said in my first posting that I wanted to include only "good" paths (HOURLY or better). I grant lsuc's connection to most of the sites you name is DIRECT. However, parkridge lists lsuc as EVENING, hcr lists it as WEEKLY, maccs is DAILY/4, and toram and sickkids don't list lsuc at all. You see my dilemma. You may say this just proves your point (and it does :-)), but I think there is *some* method in my madness. I should possibly qualify "good" as meaning "HOURLY or better in both directions". Oh and I did include your connection to ncraer. Take heart :-). ----------------------------------------------------------------------------- Don McKillican seismo!mnetor!genat!clunk!don {utcs,utai,utzoo,watmath}!lsuc!clunk!don
dave@lsuc.UUCP (David Sherman) (01/22/87)
>> Hmm. Thanks, Don. I didn't even suggest anyone try doing >> a UUCP map, because (1) pathalias can give routings and (2) >> the map would be so complex. Your attempt is nice, but is >> (just for example) missing lsuc's connections to parkridge, >> hcr, nrcaer, toram, maccs and sickkids, > >First of all, both your points are perfectly correct. But I think there >is room for another point of view, too. Pathalias is okay, but there are >a lot of sites which don't have it, don't use it, don't keep it up to date Incidentally, if you don't have pathalias or can't run it because of machine size, a neighbour may be able to help. mnetor very kindly runs "pathalias -l lsuc" once a week for us and uucp's the result over. (We have "only" 1Mb of memory, which means we haven't been able to run pathalias for over a year now.) >or any combination of the above. And there are even more users who don't >know how to use it. And I personally happen to like a pictorial approach >better. Besides, if pathalias tries to send me through lsuc, and I happen >to know that lsuc is (perish the thought!) down, then I can look on the >map for an alternate path. Pathalias isn't good at that. You're right about that. >As to the sites you name, I actually had a reason for all of them. You >will recall I said in my first posting that I wanted to include only "good" >paths (HOURLY or better). I grant lsuc's connection to most of the sites >you name is DIRECT. However, parkridge lists lsuc as EVENING, hcr lists >it as WEEKLY, maccs is DAILY/4, and toram and sickkids don't list lsuc at >all. You see my dilemma. You may say this just proves your point (and it >does :-)), but I think there is *some* method in my madness. I should >possibly qualify "good" as meaning "HOURLY or better in both directions". Aha. Interesting question. If we poll a given site every hour (subject to dialer availability), which we do for all our local sites, and if that site polls us once a day PLUS calls us whenever it has traffic, what should be the weighting? There's a problem built in to the pathalias weighting system, and I don't know of a solution. Most systems will call their mail neighbours on demand, except for some long-distance links which are restricted to evenings. So if A calls B daily plus on demand, is that DIRECT (local calls, according to the mod.map README) or something worse than the case where A calls B hourly plus on demand? I think machines that have dialers available should normally list their local connections as DIRECT; but they don't always do that. So, Don, you picking up the gauntlet and doing a Usenet map of Toronto too? :-] David Sherman -- { ihnp4!utzoo seismo!mnetor utai watmath decvax!utcsri } !lsuc!dave
tim@dciem.UUCP (Tim Pointing) (01/23/87)
Fairly recently somebody (campbell@maynard.BSW.COM / maynard!campbell) posted
a version of pathalias that runs on machines as small as the lowly PDP-11.
This is actually the most recent version of pathalias (Jan '86) converted
to use disk files rather than great gobs of memory. It *does* work (as
I can testify to since "dciem" is a PDP-11/70) but is really slow (something
that you really probably want ot let run overnight.) If anybody in the
Toronto area want a copy of it, let me know; if there is sufficient demand
(however much that is!), I'll post it to net.sources with a tor distribution.
Tim Pointing, DCIEM
{utzoo,utzoo,utcsri}!dciem!tim
P.S. Since the total size is ~100Kb, the threshold for posting the source
will be quite high.
% ls -s
total 217 15 addnode.c 5 makedb.c 18 pathalias.1
3 CHANGES 5 config.h 14 mapaux.c 16 printit
4 Makefile 10 def.h 27 mapit.c 12 printit.c
2 NOTES 2 extern.c 5 mem.c 10 putnode.c
4 README 3 local.c 31 parse.c 1 y.tab.h
8 addlink.c 6 main.c 16 parse.y
--
Tim Pointing, DCIEM
{decvax|ihnp4|watmath}!utzoo!dciem!tim
or uw-beaver!utcsri!dciem!tim
or seismo!mnetor!lsuc!dciem!tim