W3C home > Mailing lists > Public > www-talk@w3.org > March to April 2003

Re: comments? mirrors.txt

From: Andre John Mas <ajmas@newtradetech.com>
Date: Mon, 31 Mar 2003 14:57:49 -0500
Message-ID: <3E889DBD.3070507@newtradetech.com>
To: Justin Chapweske <justin@chapweske.com>
CC: www-talk@w3.org

Thanks for the references. I have taken a look through the
documentation and have a few comments:

Traditionally mirrors provide a reference to their geographic
location. While by no means a guarantee to improved transfer
speeds, this is a useful reference, especially for clients that
wish to use a single source for data transfers (as opposed
to parallel downloads). Also a brief 'advertisment' of the name
of the hosting service is provided. While this information is
not necessary for accessing the data, I believe that this
information should be provided on an optional basis. Some
companies hosting the mirrors do want to have their name
mentioned as for them it advertisment in exchange for service.

Update frequency is also of use as this allows the client
to decide which site to visit based on age of the document.

The reason I suggest a grid-reference for geographic locality
is the this is language neutral and can easily be parsed and
used to estimate which mirror is likely to best based on
distance, without having to speak to each one. The geographic 
information could be added be component of the X-URI-RES
header (though this would necessitate a change in the spec):

X-URI-RES: <service uri> ; <service type> [; target uri] [; grid-ref]

The 'detailed' mirror information could be recognized fields in
the data file referred to in the case N2L and N2Ls.

BTW Part of the idea of the mirrors.txt file is that it is human-
readable, computer parsable and easy to retrofit to existing browsers.


Justin Chapweske wrote:
> The only thing that I don't like about this is that normal HTTP 
> mirroring is very insecure.
> Our work with the "Content-Addressable Web" uses secure checksums and 
> some HTTP extensions to provide an alternate way of solving the mirror 
> problem.  Our paper on "HTTP Extensions for a Content-Addressable Web" 
> can be found at 
> (http://open-content.net/specs/draft-jchapweske-caw-03.html).
> You may also be interested in its companion specification, "The Tree 
> Hash EXchange format (THEX)" at 
> (http://open-content.net/specs/draft-jchapweske-thex-02.html).
> We also have a very basic XML-RPC protocol for lease-based mirror 
> advertisement at (http://open-content.net/specs/).
> Also, there is a functioning Content-Addressable Web header proxy that 
> you can feel free to play around with.  Its currently used for the Open 
> Content Network and can be used as follows:
> bash$ HEAD 
> http://gw1.open-content.net:8080/gateway/head?uri=http://etree01.archive.org/etree/moe1997-03-28dnk.shnf/moe1997-03-28d1/moe1997-03-28d1t06.shn 
Received on Monday, 31 March 2003 15:00:46 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:33:04 UTC