home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!spool.mu.edu!agate!doc.ic.ac.uk!uknet!keele!nott-cs!florence.linuxnet!dpg
- From: dpg@cs.nott.ac.uk (`Grave' Dave Gymer)
- Newsgroups: alt.galactic-guide
- Subject: Re: TUG! 2 --> (The Net Guide)
- Keywords: TUG admin network gopher irc tcp-ip
- Message-ID: <C17KEw.2u6@florence>
- Date: 21 Jan 93 14:32:56 GMT
- References: <C0x499.2r3@cs.nott.ac.uk> <gavin.727489695@sorokin>
- Sender: dpg@cs.nott.ac.uk (Dave `geek' Gymer)
- Organization: Association of Slack Producing Geeks
- Lines: 59
-
- In article <gavin.727489695@sorokin> gavin@sorokin.anu.edu.au (Gavin Longmuir) writes:
- >dpg@cs.nott.ac.uk (`Grave' Dave Gymer) writes:
- [...]
- >This looks like a great idea, but I can see some problems - with a networked
- >system and site that runs a server must be able to add new articles to the
- >'database' and tell all other servers if has this new information (the info
- >should only be transferred if required). Perhaps I'm talking about some kinda
- >cross between IRC and the gopher.
-
- Well, I as thinking that each server (other than the `root' one) would at some
- point in time pull down the index from some other server or set of servers,
- see what was new or changed, and then pull down those articles into its local
- database. In making each server responsible for its own database, we also
- allow it to set its own policy about when to update stuff; each server could
- also refuse to service certain hosts at certain times, thus allowing it to set
- a policy on when it will allow other servers to update from its database.
-
- The server updating utself then appears to the server being updated from as a
- normal client process, simplifying both the protocol and the code.
-
- >Given my above idea - there would have to be a new protocol with at least 4
- >different request/information formats (packets etc..); one giving the physical
- >status of all server links,
-
- I don't think this is necessary or even desirable; there's no reason why the
- links should be anything other than transitory, created anew for each request
- (this is how gopher works).
-
- >one for the guides index (I can't any reason why
- >there can't be more than one entry for a topic - like dictionaries)
-
- Absolutely.
-
- >which also
- >contains the locations of the entries (the guide should evolve so when a server
- >is decommissioned all articles in it's database [which don't exist elsewhere]
- >are forgotten by the other servers).
-
- Well, I can see two possibilities here:
-
- 1. All servers hold all articles. This is in some repects desirable as it
- allows readers to fetch any article required from a relatively local site.
-
- 2. As stated earlier, each server could periodically query other servers as to
- the articles available, and build index and cross reference lists from the sum
- of this information, and use gopher-style links merely to the entry text
- itself.
-
- [...]
- >Just my ideas (but I may just like to work of such a beast!)
-
- Well, I sure as hell don't want to have to tackle such a thing alone! :-)
-
- -- Dave
- --
- `Grave' Dave Gymer | ``The only rule in Calvinball is that you can't
- 42 St Mary's Park | play the same rule twice.'' -- Calvin and Hobbes
- Louth, Lincs |----------------------------------------------------
- LN11 0EF, England | Warning, Linux fanatic in transit.
-