home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!cs.utexas.edu!zaphod.mps.ohio-state.edu!darwin.sura.net!europa.asd.contel.com!emory!tridom!wabwrld!willard!dawson
- From: dawson@willard.UUCP
- Newsgroups: comp.bbs.waffle
- Subject: Re: News/mail reader
- Message-ID: <sy3ouB1w165w@willard.UUCP>
- Date: Mon, 23 Nov 92 08:50:27 EST
- References: <30HmuB2w165w@eastwind.mcds.com>
- Organization: Willard's House BBS, Atlanta, GA -- +1 (404) 664 8814
- Lines: 94
-
- chorn@eastwind.mcds.com (Christopher Horn) writes:
-
- > dawson@willard.UUCP writes:
- >
- > [initial argument deleted]
- >
- > > Ok, let's have a counter example. Supposing files are sorted by creation
- > > date/time (as would apparently be the case with your technique). Suppose
- > > there were no new articles, and one was expired. The users' join files
- > > are now out of sync:
- > >
- > > Before unbatching : 1, 2, 3, 4, 5
- > > After unbatching : 1, 2, 3, 4
- >
- > Suppose that the files are not sorted by time/date, only resequenced.
- > (Renumbered might be a better term)
-
- But, if you're resequencing them as they come in, they ARE sorted by creation
- date/time... as a side-effect.
-
- >
- > Before resequence : 1, 2, 3, 4, 5
- > Dupes killed : 1, 2, 4, 5
- > After resequence : 1, 2, 3, 4
-
- If { 1, 2, 4, 5 } are dupes and are erased, then only { 3 } is left? Which
- would get resequenced to 1, as that's the highest unused number in your
- example...
-
- >
- > No join files are out of sync. I think this is what Kevin meant.
- >
- > > -----
- > > Suppose there was one new article, and one was expired. No change to the
- > > users' join files, but it now points one too high:
- > >
- > > Before unbatching : 1, 2, 3, 4, 5
- > > After unbatching : 1, 2, 3, 4, 5
- >
- > You seem to be thinking they are numbered from 1 up, which is not really
- > the case. Take the lowest numbered file in the directory as the base.
-
- Yes, I admit (again) that my arguments were based on inferences from previous
- discussion, and that I have not looked at the actual implementation of DupWeed.
-
- >
- > Before unbatching : 1, 2, 3, 4, 5
- > After resequence : 2, 3, 4, 5, 6 (as per the expire one, add one)
- >
- > This works because the lowest numbered files are also the oldest and should
- > be the first removed by an expire program (assuming my understanding of how
- > expire programs expire news...)
-
- As I just pointed out.. sorted by d/t as a side-effect.
-
- >
- > > > Sounds rather simple to me.
- > >
- > > Too simple, maybe? I think I'll just avoid re-sequencing my articles.
- >
- > Yes and no. Simple to do, and users don't wonder why articles jump from
- > 1856 to 1858, etc.. But, bad because :
- >
- > 1. The INITIAL kill dupes, resequence will screw up the pointers
- > no matter what you do. So the first time penalty is high if you
- > have many users.
- > 2. You have to kill dupes, batch for downstream sites, and then
- > do the resequence seperately for it to work right.
- >
- > Also note for it to work right when getting news from multiple sites,
- > you have to search the article list from the oldest to the newest. If
- > you kill the dupe that is older, a user may have read it, and when
- > resequencing will mess up their join file.
-
- True. My own code kills the newer copy for each dupe found. Otherwise,
- you just wind up reading the same messages again and again.. and when you
- go back to see why, the message you thought you had read is no longer
- there! Talk about confusing to the avg user...
-
- >
- > > How does re-sequencing enhance the general operation of Waffle?
- >
- > Aesthetic reasons?
- >
- > ---
- > Christopher Horn | "We're all caught in a state of decay..."
- > chorn@eastwind.mcds.com | The East Wind +1 201 875 7063
-
-
- --
- dawson@willard.UUCP (Willard Dawson)
- gatech!vdbsan!willard!dawson
- emory!slammer!willard!dawson
- Willard's House BBS, Atlanta, GA -- +1 (404) 664 8814
-