Support multibyte username on xml response.

[ksauzz]

List users and show user API doesn't support multibyte on xml response. So name is garbled.

create user

[vagrant@localhost ~]$ nkf --utf8 user.json
{"name": "ほげほげ", "email":"test@example.com"}
[vagrant@localhost ~]$ curl -i http://localhost:8080/riak-cs/user -H "Content-Type: application/json" -d @user.json
HTTP/1.1 201 Created
Vary: Accept
Server: Riak CS
Location: http://localhost:8080/riak-cs/user
Date: Mon, 29 Jul 2013 05:58:22 GMT
Content-Type: application/json
Content-Length: 263

{"email":"test@example.com","display_name":"test","name":"\u307b\u3052\u307b\u3052","key_id":"X9Z01FRHMO1YJWZT-1B7","key_secret":"ueFnnV1pwc1VNg_vHvnLKC6Ah1nw888rO90PnQ==","id":"6d0e53854eed0b7f3df6605aac08723855a89091bc84848f9346b0cfd39cd08b","status":"enabled"}[vagrant@localhost ~]$

list user

XML response

[vagrant@localhost ~]$ ./s3curl.pl --id admin -- -x localhost:8080 http://riak-cs.s3.amazonaws.com/users

--OP3SKSJCRRKG6UJN3NZ50NKLFIJ47Y3
Content-Type: application/xml

<?xml version="1.0" encoding="UTF-8"?><Users><User><Email>test@example.com</Email><DisplayName>test</DisplayName><Name>����</Name><KeyId>X9Z01FRHMO1YJWZT-1B7</KeyId><KeySecret>ueFnnV1pwc1VNg_vHvnLKC6Ah1nw888rO90PnQ==</KeySecret><Id>6d0e53854eed0b7f3df6605aac08723855a89091bc84848f9346b0cfd39cd08b</Id><Status>enabled</Status></User></Users>
--OP3SKSJCRRKG6UJN3NZ50NKLFIJ47Y3
Content-Type: application/xml

<?xml version="1.0" encoding="UTF-8"?><Users><User><Email>user@example.com</Email><DisplayName>user</DisplayName><Name>user</Name><KeyId>O0YC0RKWM69HIG0NDVG1</KeyId><KeySecret>QdHYZVitFaGqfh-LSriGcMU_NDYNAh9z2MHhXg==</KeySecret><Id>02e40edb2ad4049bb75c2d4963f720a640f9e1f716c721f326759473afbda80d</Id><Status>enabled</Status></User></Users>
--OP3SKSJCRRKG6UJN3NZ50NKLFIJ47Y3

JSON response

[vagrant@localhost ~]$ ./s3curl.pl --id admin -- -x localhost:8080 -H "Accept:application/json" http://riak-cs.s3.amazonaws.com/users

--LHDGOZNBC03AS9SJORMHHDS4WMEPTT9
Content-Type: application/json

[{"email":"test@example.com","display_name":"test","name":"\u307b\u3052\u307b\u3052","key_id":"X9Z01FRHMO1YJWZT-1B7","key_secret":"ueFnnV1pwc1VNg_vHvnLKC6Ah1nw888rO90PnQ==","id":"6d0e53854eed0b7f3df6605aac08723855a89091bc84848f9346b0cfd39cd08b","status":"enabled"}]
--LHDGOZNBC03AS9SJORMHHDS4WMEPTT9
Content-Type: application/json

[{"email":"user@example.com","display_name":"user","name":"user","key_id":"O0YC0RKWM69HIG0NDVG1","key_secret":"QdHYZVitFaGqfh-LSriGcMU_NDYNAh9z2MHhXg==","id":"02e40edb2ad4049bb75c2d4963f720a640f9e1f716c721f326759473afbda80d","status":"enabled"}]

show user

XML response.

[vagrant@localhost ~]$ ./s3curl.pl --id admin -- -x localhost:8080 http://riak-cs.s3.amazonaws.com/user/X9Z01FRHMO1YJWZT-1B7/
<?xml version="1.0" encoding="UTF-8"?><User><Email>test@example.com</Email><DisplayName>test</DisplayName><Name>����</Name><KeyId>X9Z01FRHMO1YJWZT-1B7</KeyId><KeySecret>ueFnnV1pwc1VNg_vHvnLKC6Ah1nw888rO90PnQ==</KeySecret><Id>6d0e53854eed0b7f3df6605aac08723855a89091bc84848f9346b0cfd39cd08b</Id><Status>enabled</Status></User>

show user

JSON response

[vagrant@localhost ~]$ ./s3curl.pl --id admin -- -x localhost:8080 -H "Accept:application/json" http://riak-cs.s3.amazonaws.com/user/X9Z01FRHMO1YJWZT-1B7/
{"email":"test@example.com","display_name":"test","name":"\u307b\u3052\u307b\u3052","key_id":"X9Z01FRHMO1YJWZT-1B7","key_secret":"ueFnnV1pwc1VNg_vHvnLKC6Ah1nw888rO90PnQ==","id":"6d0e53854eed0b7f3df6605aac08723855a89091bc84848f9346b0cfd39cd08b","status":"enabled"}[vagrant@localhost ~]$

[reiddraper]

Is this a 1.4 blocker?

[ksauzz]

No. Thank you for the attention.

[shino]

Good reference for Latin-1 and Unicode in Erlang.
Codepoints, encodings, string literals, binaries and more.
http://www.erlang-factory.com/conference/ErlangUserConference2013/speakers/PatrikNyblom

[reiddraper]

Thoughts on priorities of this for 1.5? Scope of work?

[kuenishi]

It's not a bit feature so I think we can do it.

[kuenishi]

I think it's too much to support unicode. But the bad situation is, users can create a user named with unicode. So adding a validation that accepts only ascii printable characters would be enough. Any thoughts? CC: @shino

[kuenishi]

As there are no customer request for now, I'll suggest above change to Riak CS and decision not to support unicode chars as user names.

[shino]

I'll suggest above change to Riak CS and decision not to support unicode chars as user names.

+1

[kuenishi]

Will be solved after #787 fixed.

[shino]

Addressed by #794

[reiddraper]

Previous comments suggested we would not_ support unicode usernames, but #794 seems to add support them. Is that correct? Are there any issues with names that might have been saved with previous versions of Riak CS?

[kuenishi]

I assume that any previous version of Riak CS cannot print multibyte usernames correctly and thus no one is using unicode characters.
My aim was to reduce complexity and further fix, but Shino's fix on #794 has unexpected side effect, which is good. We still won't recommend unicode names, but it works so far.

[shino]

Fixed by #807.
Non-ASCII UTF-8 characters can be used and printed properly.