interrupting urlconnection's outputstream

[zynasis]

hi all,

I would like to add timeout functionality to a urlconnection's outputstream write.
I believe the write's flush is blocking and causing long delays on some transmissions. i am not willing to use apache http library, or any other 3rd party library for this.

currently i am wrapping the outputstream in a seperate thread that write's, flushes and closes the stream. timeouts are implemented with a join on the thread with limit on wait of the timeout amount.

after the join returns, i then check the status of the thread's write and if it is not complete, i close the outputstream and send an interrupt to the thread.

firstly, this seems messy, but i do not know of a better way (please explain if there is).
is there a better way of interrupting the outputstream? or a better way of doing this at all?

[fishtoprecords]

Is it your code that is writing to the output stream? If so, why do you want to time it out?

I can see setting a timeout on reading from some other site.

But if you are creating the output, its your job to do it quickly and efficiently.

[zynasis]

yes, my code is doing the write.
however, there are network issues outside my control which re-connections and retries fix.

the writes can take up to 15 minutes at times and i would rather have the thing bail out and stop and retry before waiting this long

[fishtoprecords]

You need to buffer the output so its all ready at once. You should not be generating more than say ten thousand bytes, so put them in a ByteArrayOutputStream and then send it all at once.

You can, of course, just setup a thread with a timer and use it to try to interupt the output.

But you really don't want to do that. Just put the output in a buffer and send it all at once.

Your code should *not* be sending megabytes. If you need to, say send a video file, put it in an accessible place and send a URL. Let apache deliver the big stuff.

[zynasis]

it already is buffering and sending in one transmission.
the amount of data is approx 400 bytes so it isnt size problems.

i think we hav a dicky router somewhere

[zynasis]

a bit more info on the problem causing me to write this:
- data is always the same size at approx 400 bytes
- data writes & flushes are usually quick at less than a second
- sometimes a single write lags up and takes over 15 minutes, meanwhile other writes are still going quick
- i believe if we timeout the write and reconnect and send, the next attempt may be fast
- there is no synchronization or locking, all sends are sent in new threads on new instances which are independent from each other