> Does anyone have any experience with ip-less accounts?
>
> Several IPPs are offering "IP-less" accounts. To illustrate, you
>open an account for zzz.com. There is a unique IP address for zzz.com.
>They you open an IP-less accoutn (varying called virtual server, etc)
>for yyy.com. yyy.com has the same ip address as zzz.com but request to
>www.yyy.com are routed to the proper directory in the zzz.com disk
>space by a combination of software and hardware.
this sounds awfully much like the name-based virtual webserver scheme from
HTTP/1.1. it's been a standard feature in Apache since at least version
1.2.
going back about three years, httpds ran as standalone daemons on the
server, and each one listened to a specific port on a specific IP address.
the traditional port for HTTP is number 80, so if you wanted multiple
webservers on the same machine, you could either use several nonstandard
ports on the same IP (all those www.domain.com:8000 beasties we used to
see), or use multiple IPs on the same machine.
most people chose the latter, which resulted in the Great Wave of IP
Depletion. machines all over the internet suddenly had as many as several
hundred IP addresses each. some of you will probably remember the 'NIC
threatening to charge annual fees to ISPs per IP address, or all the hoopla
about IPv6, which would provide enough address space to network every
electron in the universe.
the solution which actually took hold was for httpd designers to make their
software a bit more robust, and for the protocol itself to adopt a couple
more pieces of information which would support that additional
functionality.
in older versions of HTTP, when a client wants a webpage from a server, it
was acceptable to send the message:
GET /path/to_the/data/file.html HTTP/1.0
because it was implicitly assumed that only one domain's webserver would be
listening to that port on that IP address. in effect, connecting to the
server at all determined the domain the request was for.
HTTP/1.1 changed the rules slightly, requiring a fully-qualified URI in the
page request:
GET http://www.domain.com/path/to_the/data/file.html HTTP/1.0
eliminating the assumption that the connection uniquely determined the
target domain.
with that kind of information in the request, it became a fairly simple
matter to change the httpds so they kept some sort of mapping between
domain names and implicit file paths. if there's one thing the web
encourages, it's text-based pattern matching..
the directive from an Apache config file which defines a name-based virtual
webserver is:
<VirtualHost 10.10.10.10> # <-- IP address
ServerName www.domain.com # <-- domain name
DocumentRoot /path/to_the/website # <-- local filepath
</VirtualHost>
which will take the HTTP/1.1 request shown above (assuming it comes in
through IP 10.10.10.10), and translate it into a request for the file:
/path/to_the/website/path/to_the/data/file.html
the technique was stolen more or less in bulk from inetd, the network
superserver. that listens a large number of ports simultaneously, and
launches specific handler programs based on the port the request came in
on. the practical upshot is that a single machine with a single IP
address can now host websites for an arbitrary number of domains.
every hosting service i know of uses name-based virtual hosting.. it scales
far better than setting unique IPs for each httpd.
mike stone <[EMAIL PROTECTED]> 'net geek..
been there, done that, have network, will travel.
____________________________________________________________________
--------------------------------------------------------------------
Join The Web Consultants Association : Register on our web site Now
Web Consultants Web Site : http://just4u.com/webconsultants
If you lose the instructions All subscription/unsubscribing can be done
directly from our website for all our lists.
---------------------------------------------------------------------
