Daniele Ricci opened 10 years ago
|
|
I do not see any patch attached. Did you forget to attach it? To be honest I do not understand it's purpose but maybe once it is attached I can deduct from the code. |
|
Artur Hefczyc wrote:
Don't worry, nobody has these kind of powers :) I forgot to attach the patch, sorry. |
|
Daniele, I guess you are OK with the source code disclaimer in this case as well? By not understanding purpose of the patch, I meant that the description is not clear to me. |
|
Artur Hefczyc wrote:
Absolutely.
Oh, ok :) |
|
Accepted source code disclaimer on behalf of Daniele following his comments. Wojciech, please take a look at the patch and consider including it into our source code. |
|
|
|
Don't call me a pain :) but I think this can be expanded to use ExtendedPresenceProcessorIfc. What do you think? My own need is to add a custom child to every presence, including unavailable presence coming from the roster. Do you think it's better to modify RosterAbstract.getCustomChild to return an array/list of children (I could extend RosterAbstract then) or use an ExtendedPresenceProcessorIfc? I can write a patch, but I'd like to hear your opinion first. Thanks. |
|
I did not review the code and topic well enough to comment on this. I leave a decision about it for Wojciech. |
|
I was forced to disable offline-roster-last-seen as it was messing up (see #3463) and causing additional processing due to fact it generates offline presences for every one in users roster, even that some users are online (but our server will know it that after presence probe). So this features causes for online users from roster to generate presence offline followed by presence online packets. |
|
Andrzej Wójcik wrote:
I can't reach #3463 (forbidden). What was that about? |
|
In principle what Andrzej has summarised - non-standard/against-RFC behaviour of sending unavailable presence prior to the actual presence. |
|
Daniele Ricci wrote:
I've opted to modify RosterAbstract so now I've chosen this approach because this feature is still experimental and as shown above - it breaks specification somewhat (sends two I would say it's better to utilize feature implemented in commit:aaa344b7 and levarage mentioned
|
|
Wojciech Kapcia wrote:
Very well, thanks.
I agree, I was probably too rush on writing code that way :) I will do some experiments with this new approach. Thank you! |
|
I guess I can close the ticket now, let me know if you need to reopen it. |
Type |
Patch
|
Priority |
Minor
|
Assignee | |
RedmineID |
2530
|
Version |
tigase-server-7.1.0
|
Spent time |
37h 30m
|
This patch allows the RosterFlat implementation to inject a delay element inside presence stanzas with the last seen timestamp.
This breaks APIs I'm afraid, unless we want to keep the old getCustomStatus method and behaviour to work together with the new getCustomChild.
For this reason, this is probably an unwanted patch, but I'm submitting it anyway because I'd like to have your feedback and eventually modify it to be included in the source tree.
0001-Allow-injection-of-custom-child-element.patch