incubator-ooo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Kay Schenk <kay.sch...@gmail.com>
Subject Re: [DISCUSS][PROPOSAL] redirection of update services to www.openoffice.org
Date Thu, 29 Mar 2012 16:34:56 GMT
On Wed, Mar 28, 2012 at 11:35 PM, Oliver-Rainer Wittmann <
orwittmann@googlemail.com> wrote:

> Hi,
>
>
> On 28.03.2012 18:43, Joe Schaefer wrote:
>
>> Well I wouldn't say it like that Kay. The problem
>> with any update service is the sheer number of clients
>> out there configured to abuse it.  There are a number
>> of options available, but most of them revolve around
>> providing an Apache C module to at least cut down on
>> the redundant traffic before showing it to your script.
>>
>> Is it actually a cgi script at this point that you
>> are trying to service the traffic with, or just a static
>> file?  Right now our webserver is configured to disallow
>> any attempts to POST data to us, so none of yesterday's
>> traffic was handled by any of your work in this regard-
>> it was all interrupted by the server with a "Method Not
>> Allowed" 4xx response.
>>
>> I'd be willing to try experimenting again if in fact
>> you are trying to service that traffic with a static
>> file that naturally ignores POST data instead of with a
>> cgi script.  The point is to figure out how to see if
>>
>> those misconfigured clients will back-off if given
>> an expected response.
>>
>>
>>
> I am not sure about which clients you are talking about.
> If former OpenOffice.org instances are meant I can share here what I have
> learned during the integration of serf as the new HTTP/WebDAV client
> library in AOO:
> When OOo wants to load a resource (file or folder) via HTTP/HTTPS it sends
> a PROPFIND request in order to find out, if the corresponding server
> "talks" WebDAV. May be the request is sent several times in order to
> workaround an unreliable network. A HEAD request might follow. Then OOo
> sends the GET request (in case of a file) to load it.
> These HTTP requests should be observed on the server providing the update
> service when OOo automatically or the user manually triggers a "check for
> updates".
>
> Best regards, Oliver.
>

Oliver--

This is interesting. Can it be applied to the older versions already
installed by users? And, is so, can you supply details, or a "proof of
concept" installation?

Day before yesterday, we  tried to do a test (DNS re-routing) with the
older "update.services.openoffice.org" (I don't even know which version
this was the update service for), I am wondering if those *older* clients
were doing something different from what say the newer 3.x versions are.
And yep, there were a LOT of them! This is what caused the web server issue.

I did indeed see all the "POST" errors (disallowed by the webserver) that
Joe is referring to. I did several tests yesterday with my re-routed
update36.services.openoffice.org and things worked pretty well. I'm hoping
I can contact you later today about this and have you do an update via
Windows. I need to track down the web logs from what *I* did yesterday to
see what they reveal.

What I've done now is, instead of just putting the "null" update feed out
there is a feed which points to a "page" URL so folks can chose an update
to install -- currently it's going to the developer page on the wiki.

I will tailor it more to see if I can make it more "generic" (absence of
language and platform), with just a buildid, and see if it still works. I
think an approach like this will be more "friendly" for existing users.

Later ....

-- 
----------------------------------------------------------------------------------------
MzK

"Women and cats will do as they please,
 and men and dogs should relax and get used to the idea."
                                                                    --
Robert Heinlein

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message