Le 2 décembre 2006 à 17:10, James Edward G. II a écrit :
On Dec 2, 2006, at 5:50 AM, F. Senault wrote:
Le 2 décembre 2006 à 11:16, F. Senault a écrit :
Forgive my weak knowledge of email types here, but if someone sent a
short message to the list with an attached Ruby script, would it be a
multipart message? If so, that would be rejected by the NNTP host?
Yes, but it won’t be multipart/alternative (which means different
representations of the same data, nearly universally HTML + plain text),
more likely multipart/mixed (simply different parts).
For instance, with a quick search, I found this one :
Yes, but it won’t be multipart/alternative (which means different
Content-Type: multipart/mixed; boundary=“sm4nu43k4a2Rpi4c”
Am I understanding right Fred that your filters do allow multipart/
mixed?
I ask because I cannot find that the following message was moved to
Usenet:
On Sun, 2006-12-03 at 01:10 +0900, James Edward G. II wrote:
Forgive my weak knowledge of email types here, but if someone sent a
short message to the list with an attached Ruby script, would it be a
multipart message? If so, that would be rejected by the NNTP host?
I have a feeling this is what is, or at least was, happening - I wasn’t
going to say anything (since it was prior to December first) but I
noticed that posts with attachments never made it across.
That’s not systematic.
ruby-talk 226884, for example, doesn’t appear to have made it to the
newsgroup.
If you have an e-mail version of the message, could you send it to me
with all headers as a (g)zip file ? I’ll run it manually through the
filters and look at what could cause the rejection. (Maybe the base64
encoding in this one, passing for a binary post.)
Le 2 décembre 2006 à 18:05, James Edward G. II a écrit :
Am I understanding right Fred that your filters do allow multipart/
mixed?
I ask because I cannot find that the following message was moved to
Usenet:
Okay, it was interpreted as binary (probably the base 64 part). I’ve
lifted that condition (I moved clr to the binaries allowed groups) but
messages like this will probably be plagued by a bad distribution, maybe
raising even more difficult to troubleshoot problems.
On Sat, 02 Dec 2006 19:02:34 -0000, F. Senault [email protected] wrote:
messages like this will probably be plagued by a bad distribution, maybe
raising even more difficult to troubleshoot problems.
Assuming that the base64 encoding was done to cope with UTF8 in the
source, is there no way we can improve the distribution situation? The
vagaries of NNTP are not my strong point (by any means) but it strikes
me
that UTF8 will be fairly common in messages and attachments on
ruby-talk.
Did the old gateway address this somehow?
Btw. just wanted to thank both you and James for the time and resources
spent on the gateway - it’s a valuable service, and we’re all lucky to
have you working on it.
No. The new code does everything the old one did and more. The main
difference is that I can understand the code now.
Before I took over the Gateway though, it did operate through a
different NNTP host. I don’t know what that host allowed, but the
evidence suggests it may have been pretty accepting.
Le 2 décembre 2006 à 22:11, James Edward G. II a écrit :
message. Attachments are trickier and I guess we would have to
inline them. That would work for simple Ruby scripts, but it gets
more complicated when someone posts something like a zipped archive
file.
This path leads to the Dark Side ™ : content mangling of a message
(even if it’s transcoding). It’s usually considered very bad mojo to
touch anything else than the headers…
(At least, I’ll have someone to blame if I typo something in a post…
:P)
There would still be situations it couldn’t handle, but maybe we
could reduce them. This sure sounds like work though.
Yep. I’m not sure there are any good fast libraries to treat MIME, and
transfer-encodings out there ?
I’m considering making the Gateway code public now that I have
rewritten it. Maybe this will encourage enterprising souls to hack
on it a bit for features like this."
That’s definitely a good idea.
We just need to remember that we are joining two worlds with
different rules here.
Assuming that the base64 encoding was done to cope with UTF8 in the
source, is there no way we can improve the distribution situation?
Here’s what I said to Fred on the subject:
"I’m guessing we could get radical and read incoming emails with an
email library, then use that to compose a sensical Usenet post. That
way we could pull the text section of a multipart/alternative
message. Attachments are trickier and I guess we would have to
inline them. That would work for simple Ruby scripts, but it gets
more complicated when someone posts something like a zipped archive
file.
There would still be situations it couldn’t handle, but maybe we
could reduce them. This sure sounds like work though.
I’m considering making the Gateway code public now that I have
rewritten it. Maybe this will encourage enterprising souls to hack
on it a bit for features like this."
We just need to remember that we are joining two worlds with
different rules here.
Le 2 décembre 2006 à 22:11, James Edward G. II a écrit :
I’m considering making the Gateway code public now that I have
rewritten it. Maybe this will encourage enterprising souls to hack
on it a bit for features like this."
