- From: Ruediger Volk <rv@deins.informatik.uni-dortmund.de>
- Date: Sat, 11 Feb 1995 23:31:35 +0100
- To: raisch@internet.com
- Cc: uri@bunyip.com, http-wg@cuckoo.hpl.hp.com, connolly@hal.com (Dan Connolly), Markus Stumpf <stumpf@informatik.tu-muenchen.de>
Rob, to Markus Stumpf <stumpf@informatik.tu-muenchen.de> > > Btw: There is another solution to this problem. > > ifconfig supports on most platforms (at least with some minor patches) > > "alias names". Thus your host can have more than one IP address > > and you can start different httpds for different aliases. you answered > This is not an acceptable solution. > > Assigning more than one IP address to hosts (which are not gateways > or routers between different physical networks) is an irresponsible > use of a consumable resource. IMHO. I appreciate the concern very much (and while running a local registry I worked a lot to help avoid wasting of both address space and routing table size), but there are cases in which assigning multiple addresses are perfectly OK; for example - if you actually would otherwise pay the price to run an additional interface or box (well, first of all I would think that multi-homed hosts OK even if they don't forward packets between different physical networks) - if someone is using private address space (RFC 1597 - do I need my asbestos suit?) and adds unique addresses for hosts that are externally visible (in particular if the consumed unique address space is densely populated, which can be achieved e.g. by distributing host routes to all routers on the internal paths to the public network) - if the address space used for secondary addresses is densely populated and routes aggregate well - temporarily (e.g. to help while renumbering... - and note: if only a small percentage of all your hosts gets external connectivety, renumbering the externally visible secondary addresses can hurt much less than the traditional renumber all of the primary addresses game) I apologize for abusing the URI and HTTP lists with off-topic traffic; however I think the question raised by Rob is important enough to have a reply distributed to the same lists. For discussing this matter in more detail follow up is suggested to the CIDRD list (whatever it's today). Cheers, Ruediger --- Ruediger Volk Universitaet Dortmund, Informatik IRB D-44221 Dortmund, Germany E-Mail: rv@Informatik.Uni-Dortmund.DE Phone: +49 231 755 4760 Fax: +49 231 755 2386
Received on Saturday, 11 February 1995 17:32:09 UTC