eastick@mindy.uucp (Doug Eastick) (07/16/90)
Two questions. Actually just one question and some babble.
1) Has anyone used news software to distribute binay files? Say,
compressed tar files? Maybe a final header line could declare
Filetype: binary
and then go from there. My guess is that it would take WAAYY too much
work to get everyone to cooperate, but I'm thinking of an isolated
network (proprietary information).
2) Has anyone hacked news software to provide a REALLY distributed
database? I'm thinking of large corporations where there are several
manufacturing sites and several research sites.
Message-ID's could be used a product-numbers and owner
identification (e.g. <sun-ss1+@west.rnd.sun.com> which is fictitous),
RCS can be used to check-in and -out product drawings using ihave-sendme
and/or NNTP. When products are superceded, the Supercedes: line could
be used to archive the old product info somewhere. Each site would have
an index (modified history?) of what drawings/models are at each
site (contained in the message id).
If a dwg/model is requested at site A and it isn't on a local disk, the
index is searched and a SENDME is sent to site B which has the
drawing/model. If the requestor has the authority, the design files are
checked out and locked ("ci -l files") until he/she returns them.
NNTP would be used over a corporate lan and response for files should be
under a few minutes.
Any comments welcome.
--
Doug Eastick <eastick@mindy.uucp>
"He has a genetic defect and actually WANTS to go back to Sudbury" -B.E./90henry@zoo.toronto.edu (Henry Spencer) (07/17/90)
In article <1990Jul16.043412.15157@mindy.uucp> eastick@mindy.uucp (Doug Eastick) writes: >1) Has anyone used news software to distribute binay files? Say, > compressed tar files? ... The odds are that a lot of current software will mishandle it. I think C News will deal with it properly, though, if the underlying communications links can. (Don't try it over Bitnet! :-)) >2) Has anyone hacked news software to provide a REALLY distributed > database? ... Actually, the uucp maps more or less constitute one now. > Message-ID's could be used a product-numbers and owner > identification (e.g. <sun-ss1+@west.rnd.sun.com> which is fictitous), Nope nope. Message-IDs must be unique per message; a message with a message-ID that duplicates an earlier one will not even appear on your system unless the earlier one is sufficiently dead and gone. This is vital for loop prevention in the presence of redundant connections (and if you are using news for something important, you want redundant connections). But there is no fundamental problem in using the news system as a transport medium for database updates, other than the lack of end-to-end reliability guarantees. Just put your product/owner info in an "X-Somethingorother:" header or in the body of the message. -- NFS: all the nice semantics of MSDOS, | Henry Spencer at U of Toronto Zoology and its performance and security too. | henry@zoo.toronto.edu utzoo!henry