Subject:
|
Nasty 'Reference' header chopping
|
Newsgroups:
|
lugnet.admin.general
|
Date:
|
Fri, 15 Jan 1999 01:57:36 GMT
|
Viewed:
|
1365 times
|
| |
|
|
Hmm... I just noticed something alarming in the 'References' header in Tom
McDonald's recent post <lugnet.admin.general:823>. The last message ID is
truncated after the first 3 characters...(not Tom's fault)...
-------------------------------------------------------------------------------
References: <slrn79ka6i.5fg.cjc@VADER.NS.UTK.EDU> <slrn79kspj.2ho.mattdm@jadzia
.bu.edu> <369A7B28.846ABFD8@voyager.net> <slrn79l5t9.ke2.mattdm@jadzia.bu.edu>
<slrn79linv.5h1.cjc@VADER.NS.UTK.EDU> <slrn79lkoj.q60.mattdm@jadzia.bu.edu> <sl
rn79mjjn.5lb.cjc@VADER.NS.UTK.EDU> <slrn79mlv1.2ur.mattdm@jadzia.bu.edu> <369c2
a21.19719789@lugnet.com> <369C62BD.D47FB682@EarthLink.Net> <F5IyEK.8K4@lugnet.c
om> <369d88ee.21601198@lugnet.com> <F5K9D4.JGF@lugnet.com> <369e6a6a.1859138@lu
gnet.com> <F5KLoH.Cwo@lugnet.com> <36
^^^
-------------------------------------------------------------------------------
(Line has been wrapped manually at 79 columns here for readability.)
This problem was visible because Tom's message showed up as a sibling-node of
the article he was replying to, rather than a child-node of it.
I thought the truncation might be because it was posted from the web interface
where the 'References' field might've been getting chopped, because it's passed
in a hidden HTML form field. But then I posted a follow up similar to Tom's,
using Free Agent, and it did the same exact thing.
Turns out, in fact, that the length of the line is exactly 512 characters after
truncation (including the end-of-line character).
So then it looked like a server chopping problem. But I posted the same message
to lugnet.off-topic.foofoo (a non-existant group) and looked in the dead-
articles file -- and discovered that Free Agent in fact chopped the line before
submitting it via NNTP -- not the newsserver. I hand-edited the References line
to re-add the correct final message-ID, then tweaked the group name to
lugnet.off-topic.test, and resubmitted it with 'inews -h' and the server took it
fine and didn't munge it up. Here it is:
http://www.lugnet.com/news/display.cgi?lugnet.off-topic.test:41
Dang, this is frustrating.
Of all references to chop, the ones at the end of the lines are the worst ones
to chop. :-/ This means it's impossible ever to go back and repair all the
broken References lines without human assistance on a case-by-case basis. (Not
that it's important to go back and do that, but it irks me on principle.)
With regard to the web interface, I suppose it shouldn't be passing the
references in a hidden field -- it should pass the (single) parent article ID
only, and then look up the rest of the references just prior to injecting the
article. Just doens't make sense though -- Netscape 4.0 and 4.5 both limiting a
hidden field to 512 characters?! I must be missing something. I really need
hidden fields to work with big chunks like 16K for upcoming stuff. Grr.
Putting this on the back burner.
--Todd
p.s. This same problem happend to Mike Stanley in early October with slrn+Pico,
which was chopping lines at 255 characters. But here, Tom's message was posted
through the web interface, and there's no excuse for it flubbing things up.
|
|
Message has 2 Replies:
 | | Re: Nasty 'Reference' header chopping
|
| (...) <snip> Todd, once again, let me say how thankful I am that someone who knows what he's doing (or least 97% of what he doing(1)) is taking care of little blood- sucking lymed forest ticks like that that inhabit a place that I really like. (...) (26 years ago, 15-Jan-99, to lugnet.admin.general)
| | | Re: Nasty 'Reference' header chopping
|
| Todd Lehman wrote in message <369e97d3.13486202@l...et.com>... (...) after (...) This makes sense... somewhere in the RFC977bis draft it says that the maximum length of a line going over the NNTP protocol is 512 bytes. The client picked the wrong (...) (26 years ago, 15-Jan-99, to lugnet.admin.general)
|
4 Messages in This Thread:
  
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
