Hi,

I started working again on #3373:
http://projects.reductivelabs.com/issues/3373

I'd like to discuss here how to achieve the file client streaming (ie
never fully buffer the file content).

What makes thing real complex is that the response is valid only in the
block of the network request:

<pseudo-code>

http.get(...) do |response|
  ... here response is accessible by chunk
end

... response is not accessible unless the request was buffered in RAM
... which is not what we want to achieve.
</pseudo-code>

In the current version, the file content indirector REST call is done
when something tries to access the File source property content
attribute. This is called by the content property at the time of syncing
the type.

So, we either need to reverse how the content writing works (passing
done a block), or use dirty thread tricks like I did to defer the http
request to when we'll actually write content to disk, because we have to
pass a block (which will be called with the response chunk) that will
have to write the content to disk.

My current implementation solves the issue by firing a new thread that
waits the file type is ready to write something to the disk and then
handles response chunk to the writer thread. It works, but is quite
dirty (and I'm sure everybody was offensed by this hardcore code to the
point everybody was speechless :-)). For reference see DeferredResponse
in [1]

So, I'm seeking advice on how to write this code better, and will
welcome any comments or ideas.

If you need a reference to the actual patch see here:
[1] http://github.com/masterzen/puppet/tree/tickets/master/3373
-- 
Brice Figureau
My Blog: http://www.masterzen.fr/

-- 
You received this message because you are subscribed to the Google Groups 
"Puppet Developers" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to 
[email protected].
For more options, visit this group at 
http://groups.google.com/group/puppet-dev?hl=en.

Reply via email to