Allow for the TCPServer class to act like a ACE ACE_TP_Reactor

[pcunite]

It has come to my attention that the TCPServer class may not scale very well (1000+ clients). The ACE library has a class called ACE_TP_Reactor that follows the __Leader-Follower__ model. Read link below for description. What is necessary to get scalability into Poco? This is a major deterrent to the acceptance of Poco.

http://www.orbzone.org/wp-print.php?p=56

[alex]

> What is necessary to get scalability into Poco?

As I have already mentioned in [url=http://pocoproject.org/poco/wiki/tiki-view_forum_thread.php?comments_parentId=1322&topics_offset=2&forumId=6|this thread], it is wrapping http://monkey.org/~provos/libevent/|libevent] or similar library. Leader-follower applies multiple threads to handling events and saves the overhead of thread creation, but the overhead of running threads remains. For solutions to this problem, see [http://www.kegel.com/c10k.html|C10K] or [http://www.cs.wustl.edu/~schmidt/PDF/proactor.pdf|Proactor pattern] or [http://www.boost.org/doc/libs/1_37_0/doc/html/boost_asio.html]boost::asio[/url for details.

All that being said, we'd all like to see scalability improved in POCO. Unfortunately, unlike our desires, means are limited. Bottom line - we'll have to wait until someone has an itch strong enough.

[pcunite]

All that being said, we'd all like to see scalability improved in POCO. Unfortunately, unlike our desires, means are limited. Bottom line - we'll have to wait until someone has an itch strong enough.

I understand. Why not start a special page on your site to get support for this. My company can donate $1,000 USD to the cause but need to have some guarantee it would be done by a certain time. On this page let donors pledge the funds to a needed amount. When that amount is reached hire someone to add the functionality. Poco is worthless without this ability. The design goal of Poco is ''network-centric, portable applications''. The network today consists of many users. A clean API is a very worthy goal and we applaud those behind Poco but there needs to be some meat behind it.

[alex]

> I understand. Why not start a special page on your site to get support for this. My company can donate $1,000 USD to the cause but need to have some guarantee it would be done by a certain time. On this page let donors pledge the funds to a needed amount. When that amount is reached hire someone to add the functionality.

Let's use this thread as a starting point, so if anyone is wiling to pledge funds for this particular purpose, they can follow your example and post here for the time being.

In the meantime, I'll coordinate with Applied Informatics folks on details of how to go about doing what you are your proposing. As you know, I have [url=http://pocoproject.org/poco/blog/?p=150|proposed something similar recently]. We also have http://pocoproject.org/poco/info/donations.html]donations page[/url, but the response so far was underwhelming and we did not really feel there was grounds to seriously pursuing anything there for the time being. But that may not be a case, we'll see how this initiative goes.

Poco is worthless without this ability. The design goal of Poco is ''network-centric, portable applications''. The network today consists of many users.

Scalability is a worthy goal and we'd like to have the best possible. Keep in mind, though, that the company behind it is not in business of providing world-class web server, that's why we have what we currently have.

A clean API is a very worthy goal and we applaud those behind Poco but there needs to be some meat behind it.

I'm all for it. We got the initial impulse and now it remains to be seen how this initiative goes.