140472 - DNS: No escape for non ASCII host name

Status: RESOLVED WONTFIX

(Core :: Networking, defect, P3)

Description

[nhottanscp]

Mozilla escapes entire host name if it contains non ASCII.
e.g.
www.á.com -> "%77%77%77%2E%E3%81%82%2E%63%6F%6D" 

Encoding host name may be done at client resolver or by the server. That would
not work if Mozilla escapes the host name.
E.g. mdnkit by JPNIC
http://www.nic.ad.jp/ja/idn/mdnkit/download/documents/mdnkit-2.4-doc/en/spec/spec.html

I think this is important because Mozilla does not provide IDN encoding as
default and the user may need to depend on other solutions to enable IDN.

Comment 1

[Katsuhiko Momoi]

CC'ing edmon. 
I want to nominate this for nsbeta1. We would like to
claim support for multilingal domain names for the next
client release. If we escape the domain name part of it,
our support only has an incomlete story.

Comment 2

[Darin Fisher]

sounds like a good plan to me.

Comment 3

[Darin Fisher]

no time for 1.0

Comment 4

[Darin Fisher]

Attachment: patch

this patch doesn't solve the bug exactly... it just fixes the escaping so that
only non-ASCII characters are escaped.	escaping all characters was never
intended.

Comment 5

[Doug Turner (:dougt)]

Comment on attachment 84876 [details] [diff] [review]
patch

r=dougt

Comment 6

[rpotts (gone)]

Comment on attachment 84876 [details] [diff] [review]
patch

sr=[e-mail]

Comment 7

[Darin Fisher]

landed attachment 84876 [details] [diff] [review] on the trunk.

Comment 8

[Darin Fisher]

nhotta: i'm on vacation now... if this still needs to be fixed by RTM then it'll
need a new owner.

-> 1.2alpha

Comment 9

[nhottanscp]

Reassign to nhotta and i18n group will evaluate this for rtm.

Comment 10

[Frank Tang]

I think the idns is still not mature enough and it is too risky to change in
this level of code right now. Let's nsbeta1- it and push for trunk .

Comment 11

[nhottanscp]

Return this to darin.
Can we unescape non ASCII just before we send the host name to the server?

Comment 12

[nhottanscp]

*** Bug 161432 has been marked as a duplicate of this bug. ***

Comment 13

[Darin Fisher]

-> future

Comment 14

[benc]

If this is IDN, can I get the IDN QA to take this bug from me?

Comment 15

[Aaron Swartz]

For simple test cases, visit http://pi.cr.yp.to/

If this were working, you could visit http://π.cr.yp.to/ (i.e. with a greek Pi)
but instead Mozilla goes crazy. 

This should be simple to fix: just stop escaping the domain name.

Comment 16

[Simon Montagu :smontagu]

very important to fix for international, since non ascii characters are
increasingly being used in domain names; changed to nsbeta1+; assigned to nhotta
-dassi

Comment 17

[Samir Gehani]

i18n triage team: nsbeta1+/adt3

Comment 18

[nhottanscp]

IDN (Internatioanl Domain Name) is now official (bug 112979). That means sending
a domain name as raw UTF-8 is not legal. The current behavior to escape a domain
name when IDN is not enable is the right behavior.
Removing nsbeta1+. Shall I mark this as wontfix?

Comment 19

[Darin Fisher]

yeah, i think that makes sense.  marking WONTFIX.

Comment 20

[benc]

So, the initally described behavior, (escape entire hostname) is the correct
behavior if we are not in an IDN-enabled mode?

Comment 21

[nhottanscp]

Yes, if IDN is not enable, apply escaping per RFC 2396.