home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!utcsri!torn!nott!cunews!revcan!ecicrl!clewis
- From: clewis@ferret.ocunix.on.ca (Chris Lewis)
- Newsgroups: comp.mail.uucp
- Subject: Re: Are costs relevant ? (was Re: UUCP map entries)
- Message-ID: <4001@ecicrl.ocunix.on.ca>
- Date: 17 Nov 92 06:45:03 GMT
- References: <3983@ecicrl.ocunix.on.ca> <1dtdonINN58c@grasp1.univ-lyon1.fr>
- Organization: Elegant Communications Inc., Ottawa, Canada
- Lines: 65
-
- In article <1dtdonINN58c@grasp1.univ-lyon1.fr> Christophe.Wolfhugel@grasp.insa-lyon.fr (Christophe Wolfhugel) writes:
- >In article <3983@ecicrl.ocunix.on.ca> clewis@ferret.ocunix.on.ca (Chris Lewis) writes:
- >>On the contrary, they're *entirely* relevant.
-
- >I'm not convinced anymore on that point. In the design, costs
- >are definitely a good stuff, but seeing how they are more
- >and more used it does not reflect anymore the reality.
-
- They may not reflect reality, but this doesn't mean that they're irrelevant.
- In fact, the problem is that they're relevant, but wrong. Nor would
- any improvement be seen by eliminating costs. The real improvement
- is by getting people to fix their maps, as well as ensure that they're
- up-to-date.
-
- >First point: should costs reflect reall costs, delays or a combination of
- >both of those parameters ?
-
- >A cheap way may take several days whereas a not much more expensive one
- >takes a few hours. I recently had the case with a site in the usa.tx maps
- >who had a pathalias route cheap and efficient.
-
- The best way to calculate them is to first determine how often you
- connect to your neighbors. This step is the easy one. Take that
- as a starting point. Then, factor in:
- - link speed (for small + or - increments)
- - desirability of routing other people's mail thru the
- link. If you don't want other people to use the link, but
- you should be able to, then don't advertise it, just
- do it in your local maps.
- - reliability - if you go for months at a time only
- being able to reach them once a week, then say WEEK.
- Not DEMAND because it's supposed to be, but WEEK
- because that's how the link sometimes behaves.
- Take worst-case.
-
- There are really no hard-and-fast rules. There really can't be.
-
- >And...
-
- >... with the recent set of maps posted for Texas, many links have
- >changed, and in fact new ones have been added. From a two-hops
- >away from the Internet I switched to another two hops away via
- >site ks.uucp (at IBM Austin). Too bad, this site is misconfigured and
- >all mail get bounced. I declared the site as dead and restarted the
- >pathalias. News entry point: another IP site in Texas and 3 or
- >4 hops to go to my correspondant. Another link dead, and that
- >still for two or three more turns gave my unefficient routes
- >with low cost...
-
- Try sending mail to the postmasters, my experience is that it often
- helps to unjam links, or at least to get the map entries fixed.
-
- On the other hand, I often find it more reliable to only run with
- local maps (eg: province or state), and select a well administered
- machine as smart-host. I get virtually no bounces from mnie.
-
- >I'm also conscient that changing the current cost scheme would
- >be nearly impossible.
-
- To what? The *real* problem is SAs not keeping their maps up
- to date. Changing the costing scheme won't make any difference.
- --
- Chris Lewis; clewis@ferret.ocunix.on.ca; Phone: Canada 613 832-0541
- Psroff 3.0 info: psroff-request@ferret.ocunix.on.ca
- Ferret list: ferret-request@ferret.ocunix.on.ca
-