[thelist] Spreading sites between different hosts

Scott Dexter sgd at ti3.com
Fri Aug 25 10:26:44 CDT 2000


 
> It's when I try to think about what time penalties or other 
> problems might
> occur that my lack of knowledge is exposed. For example, does 
> the initial
> connection to a site hold that connection open so that all 
> further requests
> to that server are dealt with  faster than the initial connection?

No. The connection is open for the length of time it takes to download the
page. Then closed. subsequent connections incur the same overhead. Remember
http is stateless...

> 
> A link from one page on a site to another page on the same 
> site and same
> server can link simply by referencing the file name if the 
> new file (page)
> is in the same directory. Is this faster than if the link 
> referenced the
> full URL?

The browser converts relative links to full URL's internally; there is
effectively no difference here.

> 
> Similarly. If a link to another page in the same directory of the same
> server is referenced by the full URL, is the link appreciably 
> faster than
> if the link took you to a page on a different server.

The only difference you may experience is between the two servers
holistically; for example if one portion of the site is hosted on a T3 and
the other on an ISDN, there are inherent latencies regardless of the
content.

 
> In short, how much of a penalty is there if I spread the 
> pages of a site
> around several different host servers? What is the best way 
> to think about
> this kind of network/logistics problem?

The only problems you may encounter is the wonderful spaghetti of linking
between the two (and keeping them in check), but do the research to keep the
hosts equivalent with respect to bandwidth available. Other than that, rock
on--

sgd
--
http://thinksafely.org





More information about the thelist mailing list