[ 
https://issues.apache.org/jira/browse/TIKA-4178?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17804813#comment-17804813
 ] 

Tim Allison commented on TIKA-4178:
-----------------------------------

And, as happens with unseemly workarounds, there were several good reasons for 
this. This morning, I considered adding automatic url encoding to the 
httpfetcher, but that would have us building our own URL parser from scratch 
(this brought reality but not joy: 
https://stackoverflow.com/a/29948396/6281268).

I think that it is better to put the burden on the user for correctly encoding 
URLs. 

The challenge is that we need to decode the fetchkey for file fetchers in the 
/tika endpoint...users need to be able to send non-ascii fetchkeys. And, yet, 
as this shows, we also need to let users pass through correctly encoded URLs.

None of this is a problem for the /pipes or /async endpoints because the 
request(s) all go through json. But, trying to force the fetcher model onto the 
/tika (and /rmeta) endpoints has led to a less than ideal design.

We definitely need to document the workaround better. And, if anyone has a 
better option, please share it!

> HttpFetcher should handle encoded URLs properly
> -----------------------------------------------
>
>                 Key: TIKA-4178
>                 URL: https://issues.apache.org/jira/browse/TIKA-4178
>             Project: Tika
>          Issue Type: Bug
>          Components: tika-pipes
>            Reporter: Tim Allison
>            Priority: Major
>
> On the user list, João Domingues reported that the HttpFetcher corrupts URLS, 
> even ones that are properly encoded.
> https://lists.apache.org/thread/rlbyp4xymhsc3obrb6xsc3qq8w5n6jos



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to