weemba@beaver.ics.uci.edu (Matthew P Wiener) (09/03/89)
In article <1989Sep2.200616.8524@brutus.cs.uiuc.edu>, coolidge@brutus (John Coolidge) writes: > I also think the RFC needs >a rewrite, and at that time it should be decided if Lines: is worth >making required. If it or some equivalent is not there, a newsreader can't do a buffered article read-in and simultaneously give the reader a percentage. What should I do about this in Gnews 2.1? Include a "please wait, this article won't be buffered due to C News idiocy" message? Feh. -- -Matthew P Wiener (weemba@beaver.ics.uci.edu, weemba@math.berkeley.edu)
coolidge@brutus.cs.uiuc.edu (John Coolidge) (09/03/89)
weemba@beaver.ics.uci.edu (Matthew P Wiener) writes: >In article <1989Sep2.200616.8524@brutus.cs.uiuc.edu>, coolidge@brutus (John Coolidge) writes: >> I also think the RFC needs >>a rewrite, and at that time it should be decided if Lines: is worth >>making required. >If it or some equivalent is not there, a newsreader can't do a buffered >article read-in and simultaneously give the reader a percentage. >What should I do about this in Gnews 2.1? Include a "please wait, >this article won't be buffered due to C News idiocy" message? Feh. If you're going to put in such a message, it should be more along the lines of "please wait, news does not force enough support" or some such. IMHO, a package which adheres fully to some standard shouldn't be considered "idiotic" when it chooses not to implement a non-required part of that standard. That's what the entire point of making some things not required is! I'm also still not sure Lines: is the place to require such support. It's just as cheap, once one has already decided to retrieve an article, to do a stat() on the file and find out (far more precisely than Lines:) how big a given article is. What? NNTP doesn't offer support for doing this? Than the real problem is with NNTP support, not with Lines: at all... Of course, no newsreader that I know of currently uses the stat() idea to figure out the size of an article directly (their pager does, though, in many cases), while several use Lines:. For that reason I really encourage everyone to provide Lines:, because it really is useful information as things are currently implemented. Anyone who's running C News should turn on Lines: support ASAP. But I balk at pointing fingers or calling names at those who DON'T provide Lines: --- they've just chosen not to do something which is clearly an optional thing to do. Finally: If Lines: is to be required, I'd like to see it expanded (either by the same name or by some new name, Size: for instance) to carry wc-like information. The byte count is clearly useful, and I can envision word count information having some value at all. If it's required that we spend the time and space to provide Lines:, let's at least generate a bunch of useful information all at once. --John -------------------------------------------------------------------------- John L. Coolidge Internet:coolidge@cs.uiuc.edu UUCP:uiucdcs!coolidge Of course I don't speak for the U of I (or anyone else except myself) Copyright 1989 John L. Coolidge. Copying allowed if (and only if) attributed. to provide Lines:
jerry@olivey.olivetti.com (Jerry Aguirre) (09/06/89)
In article <1989Sep3.160552.25045@brutus.cs.uiuc.edu> coolidge@cs.uiuc.edu (John L. Coolidge) writes: >Finally: If Lines: is to be required, I'd like to see it expanded (either >by the same name or by some new name, Size: for instance) to carry wc-like >information. The byte count is clearly useful, and I can envision word >count information having some value at all. If it's required that we spend >the time and space to provide Lines:, let's at least generate a bunch of >useful information all at once. This came up before in a discussion about NNTP batching. A byte count set by the orriginating system is not going to be portable to other systems. For example, NNTP transmitts articles in "network" format. That means it terminates each line with a carriage-return and a line-feed. So a byte count would have an additional byte for each line than would be present if the article was on a typical Unix system. This is not just a transport issue; Some systems are going to store the article in different formats. Not only will the line terminators change but some systems might use fixed length records or (yuck) translate tabs into blanks. The "lines" header is not a CRC for checking whether the article has been corrupted. It is not going to help in article transfer or batching. It is useful for getting a "human" estimate of the size of an article though a count of "words" or "printing" characters would be more useful. Jerry Aguirre
coolidge@brutus.cs.uiuc.edu (John Coolidge) (09/06/89)
jerry@olivey.olivetti.com (Jerry Aguirre) writes: >In article <1989Sep3.160552.25045@brutus.cs.uiuc.edu> I write: >>Finally: If Lines: is to be required, I'd like to see it expanded (either >>by the same name or by some new name, Size: for instance) to carry wc-like >>information. The byte count is clearly useful, and I can envision word >>count information having some value at all. If it's required that we spend >>the time and space to provide Lines:, let's at least generate a bunch of >>useful information all at once. >This came up before in a discussion about NNTP batching. A byte count >set by the orriginating system is not going to be portable to other >systems. > [reasons: file system, line format, etc., deleted] Quite true, and I didn't mean to imply otherwise. That doesn't mean an estimate of it is not useful, especially if the count was actually 'characters' --- the number of user-readable characters in the message. >The "lines" header is not a CRC for checking whether the article has >been corrupted. It is not going to help in article transfer or >batching. It is useful for getting a "human" estimate of the size of an >article though a count of "words" or "printing" characters would be more >useful. Exactly. Neither Lines: or any replacement will be useful for anti- corruption, transport, etc., reasons. They're a user-information feature, similar to Keywords:, Summary:, Organization:, and the like. The time a character count is useful is when there are 500 lines in the article, 450 of which are blank. The time a line count is useful is: the same one (surprise :-)) so that you know the next 450 lines will be blank. Words could be a useful measure as well, perhaps the most useful of all. I think the overall point is: Lines:, Size:, or whatever-we-call-it:, is a very useful header. There's a good chance it ought to be a required header. Right now it isn't, and it's wrong to flame a system for not providing clearly optional features. Some sort of consensus needs to be found as to whether whatever-it-is is to be required, or not, and that needs to be encapsulated in a new standard. Until then, anarchy reigns --- and let the reader beware! --John -------------------------------------------------------------------------- John L. Coolidge Internet:coolidge@cs.uiuc.edu UUCP:uiucdcs!coolidge Of course I don't speak for the U of I (or anyone else except myself) Copyright 1989 John L. Coolidge. Copying allowed if (and only if) attributed.