Virtual Web Hosting Services

Daniel, We strongly support Peter Galbavy's position on web hosting services and that using a single ip address is not bad! The Internet's growth and success have made it crucial for any good sized company to have a presence on the Web. The alternatives are usually to install a leased line, router and web server which would usually take up a much larger subnet or even a full class C address or to use a web hosting service of some sort that uses only a single ip address for an entire company. This is clearly much more efficient that each Web site using a larger address space assignment and thereby placing greater strain on the diminishing address pool. Large companies have lots of budget and if they can't get the lower cost shared web server solution, they will opt for the higher solution. Our experience is that they will not tolerate the www.provider.xx/~yourco sort of addressing convention. Great, however for individuals. Birds of a feather like to fly together. Most of the big birds have their own www.yourco.xx registations and now the smaller ones now want the same sort of presentation. The designers of http should perhaps have thought this out or get thinking about a solution. In the meantime the rest of us have to find a way to make it work the most efficiently. Regards David Guthrie, Technocom plc, U.K.

Another aspect of virtual servers is that people who actually has tried implementing Daniel's second method (which was a CNAME from www.company.com to www.provider.com and referring to the page as http://www.company.com/company/) discovers that the most popular web browsers makes a dns lookup followed by a reverse lookup of the URL they are looking at and presents this information at the top of the page. This means that a reference to http://www.company.com/company/ will show up as http://www.provider.com/company/ on the user's screen. I can understand that from a marketing point, this is not a good thing. There has been endless discussions about this on the web developers lists, but I think this has been resolved by proposing a new HTTP header "Host:" instead of passing the full URL, since this would break older implementations of HTTP servers (~40000 installed). -Rickard -- Rickard Schoultz <schoultz@sunet.se> SUNET/NORDUnet/Ebone Operations Center KTH, S-100 44 Stockholm, Sweden Phone: +46-8-7908365 Fax: +46-8-241179

On Sat, 18 Nov 1995, Rickard Schoultz wrote:
discovers that the most popular web browsers makes a dns lookup followed by a reverse lookup of the URL they are looking at and presents this information at the top of the page. This means that a reference to http://www.company.com/company/ will show up as http://www.provider.com/company/ on the user's screen.
I don't think the client does a reverse lookup. But in the configuration of your httpd, you tell it what it's hostname is. When people make a wrong reference that requires a redirect, the server sends the configured name as part of the new URL. In general people have a tendensy of forgetting the trailing slash on URL's when it's a directory. This makes the client establish an extra TCP connection - the first one only to get a redirect. For example http://www.DK.net/nic would make the server redirect the client to http://www.DK.net/nic/ making it connect twice to get the wanted URL. I believe that's what you're seing. I do not see a problem with allocating an IP# for an extra pseudo-interface on a computer as the complete URL isn't transfered today. If we are not allowed to do it, people will get C's to do the same (in the name of the organisation wanting the service), which is much more expensive. -- Robert Martin-Legène, = EUnet Denmark = DKnet, Fruebjergvej 3, DK-2100 Kobenhavn O, +45 39 17 99 00
participants (3)
-
David Guthrie
-
Rickard Schoultz
-
Robert Martin-Legene