[ 
https://issues.apache.org/jira/browse/TS-167?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Sean Cosgrave updated TS-167:
-----------------------------

    Attachment:     (was: FetchSM.h)

> Support InkHttpFetchPage(s) in the TS API
> -----------------------------------------
>
>                 Key: TS-167
>                 URL: https://issues.apache.org/jira/browse/TS-167
>             Project: Traffic Server
>          Issue Type: New Feature
>          Components: InkAPI
>            Reporter: Sean Cosgrave
>         Attachments: fp_fetch_diff.txt
>
>
> This is the original description from Anirban:
> There are quite a few people that are attempting to write plugins that 
> download data from an origin server (or other
> HTTP sites) and then optionally modify the contents like the transformation 
> plugins allow for.
> However, the downloading of the pages is cumbersome and a repeated task as 
> this, should be wrapped in an API.
> The input into the function would be
> 1. provide the headers passed into YTS as part of the request
>    - the header should obviously include the url to fetch
> 2. provide the moment when to wake up the plugin's continuation. Options 
> include
>    0 - as soon as the request is made - no attempt is made to wait for the 
> download to happen
>    1 - as soon as the headers are downloaded
>    2 - as soon as a portion of the message is downloaded
>    3 - after the full download is complete
> 3. ability to not cache the data (optional) (only used if one would like to 
> override the data
> 4. ability to not run plugins (optional)
> The output of the function would be
> 1. for each of the results
>    - error codes that define the download status of each of the requests 
>    - IOBuffer ptr (or char*) to the results (the caller of the fxn has to 
> release references to the IOBuffers or if char* is used free the chunk) 
>    - time to download each object (not so critical)
> Note: This fxn doesnt support streaming the answer back to the requesting 
> client. 
> The fetches should be asynchronous and simultaneous.
> A possible model would be to have the user call INKHttpFetchPages which then 
> in turn calls INKHttpFetchPage per request.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

Reply via email to