[TIKA-912] Response charset encoding not declared, and depends on host OS (Windows/Linux)

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 1.1
• Component/s: server
• Labels: newbie patch

Environment

java version "1.6.0_26"
Java(TM) SE Runtime Environment (build 1.6.0_26-b03)
Java HotSpot(TM) Server VM (build 20.1-b02, mixed mode)

java version "1.6.0_31"
Java(TM) SE Runtime Environment (build 1.6.0_31-b05)
Java HotSpot(TM) Client VM (build 20.6-b01, mixed mode, sharing)

• Assignee: Unassigned
• Reporter: Graham Charters
• Created: 04/May/12 20:53
• Updated: 02/Mar/15 04:32

Description

When the response to the /tika servlet contains non-ASCII characters, Tika doesn't tell us what encoding it's using, and the encoding differs depending on which OS the server is running on.

This is a server running on Tomcat on Linux:

chris@lap-x201:~/projects/atamis-intranet/django/intranet$ curl -i -T documents/fixtures/smartquote-bullet.docx http://localhost:8080/tika/tika | hexdump -C
00000000  48 54 54 50 2f 31 2e 31  20 31 30 30 20 43 6f 6e  |HTTP/1.1 100 Con|
00000010  74 69 6e 75 65 0d 0a 0d  0a 48 54 54 50 2f 31 2e  |tinue....HTTP/1.|
00000020  31 20 32 30 30 20 4f 4b  0d 0a 53 65 72 76 65 72  |1 200 OK..Server|
00000030  3a 20 41 70 61 63 68 65  2d 43 6f 79 6f 74 65 2f  |: Apache-Coyote/|
00000040  31 2e 31 0d 0a 43 6f 6e  74 65 6e 74 2d 54 79 70  |1.1..Content-Typ|
00000050  65 3a 20 74 65 78 74 2f  70 6c 61 69 6e 0d 0a 54  |e: text/plain..T|
00000060  72 61 6e 73 66 65 72 2d  45 6e 63 6f 64 69 6e 67  |ransfer-Encoding|
00000070  3a 20 63 68 75 6e 6b 65  64 0d 0a 44 61 74 65 3a  |: chunked..Date:|
00000080  20 46 72 69 2c 20 30 34  20 4d 61 79 20 32 30 31  | Fri, 04 May 201|
00000090  32 20 31 39 3a 34 30 3a  35 34 20 47 4d 54 0d 0a  |2 19:40:54 GMT..|
000000a0  0d 0a e2 80 99 0a e2 80  a2 09 0a                 |...........|
000000ab

And this is a server running on Tomcat on Windows:

chris@lap-x201:~/projects/atamis-intranet/django/intranet$ curl -i -T documents/fixtures/smartquote-bullet.docx http://localhost:9080/tika/tika | hexdump -C
00000000  48 54 54 50 2f 31 2e 31  20 31 30 30 20 43 6f 6e  |HTTP/1.1 100 Con|
00000010  74 69 6e 75 65 0d 0a 0d  0a 48 54 54 50 2f 31 2e  |tinue....HTTP/1.|
00000020  31 20 32 30 30 20 4f 4b  0d 0a 53 65 72 76 65 72  |1 200 OK..Server|
00000030  3a 20 41 70 61 63 68 65  2d 43 6f 79 6f 74 65 2f  |: Apache-Coyote/|
00000040  31 2e 31 0d 0a 43 6f 6e  74 65 6e 74 2d 54 79 70  |1.1..Content-Typ|
00000050  65 3a 20 74 65 78 74 2f  70 6c 61 69 6e 0d 0a 54  |e: text/plain..T|
00000060  72 61 6e 73 66 65 72 2d  45 6e 63 6f 64 69 6e 67  |ransfer-Encoding|
00000070  3a 20 63 68 75 6e 6b 65  64 0d 0a 44 61 74 65 3a  |: chunked..Date:|
00000080  20 46 72 69 2c 20 30 34  20 4d 61 79 20 32 30 31  | Fri, 04 May 201|
00000090  32 20 31 39 3a 33 39 3a  35 32 20 47 4d 54 0d 0a  |2 19:39:52 GMT..|
000000a0  0d 0a 92 0a 95 09 0a                              |.......|
000000a7

As you can see, the data (last few bytes) is encoded differently. The Linux server encodes it as UTF-8, while Windows is using something strange, probably Windows-1252, where 0x92 is a curly quote and 0x95 is a bullet point.

A client can't know what encoding the server used, because the Content-Type is just text/plain with no encoding.

Ideally I would like it to use UTF-8 always, so that the client doesn't have to do extra work to decode it. The attached patch does that, and declares it.