| gabriel rosenkoetter on Thu, 27 Sep 2001 21:10:28 +0200 |
|
On Thu, Sep 27, 2001 at 10:45:11AM -0400, Kevin Brosius wrote:
> Yes, this is a problem. I was exchanging some mail with the cvsup
> maintainer recently about the possibility of a C/C++ port, but it's a
> very large task (40,000+ lines of code).
Huh. I'd rather just see a properly portable implementation of the
Modula-3 compiler. It sounds like the language is pretty useful.
> How will you address loss of traffic? If it's a private network,
> I'd guess you'll be better off, but will you allow clients some
> method of updating when they go offline for some period of time
> and miss messages?
Right now I'm trusting INN to take care of that kind of thing. If a
message is queued and the transport dies, then the history on the
remote INN host will not sync with the history on the sending
server, and they'll resolve this by transfering the message after
the feed is reestablished. At least, that's how I understand this.
I doubt it's perfect, but even if you do end up with a partial
message, then the processing process on the receiving end will
either error out in parsing the post or if, miraculously, it
doesn't, then the md5 sums won't match up and it will request a
full post of the ,v file in question.
This could mean that messages get to the anoncvs servers out of
order, but that's no big deal and is precisely the reason I'm
tagging posts with sequence numbers. (If you get a higher sequence
number than you want, you wait a spell for the right sequence
number, then either bitch loudly to the operator or to the posting
server on a separate newsgroup.)
--
~ g r @ eclipsed.net
Attachment:
pgp2039FsjpeK.pgp
|
|