above
The above arguments show that threads can perform at least as well as events for high concurrency and that there are no substantial qualitative advantages to events.
up to this point,
これまでに
Up to this point, we have largely argued that threads and events are equivalent in power and that threads can in fact perform well with high convurrency.
believe
We believe that threads provide a better programming abstraction for servers with these two properties.
For instance
For instance, many event systems “call” a method in another module by sending an event and expect a “return” from that method via a similar event mechanism.
To 動詞, 文章
To solve this problem in threaded servers, we propose a mechanism that will in threaded servers, we propose a mechanism that will enable dynamic stack growth.
In order to 動詞，文章
In order to understand the application, the programmer must mentally match these call/return pairs.
furthermore
Furthermore, these call/return pairs often require the programmer to manually save nad restore live state.
because
Freeing this state at the correct time can be extremely difficult because branches in the applications’s control flow can cause deallocation steps to be missed.
2コラム辺り2～3個。
今のところmaxで3個。
For example,
For example, our own Ninja system ended up using threads for the most complex parts, such as recovery, simply because it was nearly impossible to get correct behavior using event.
in addition
In addition, applications that didn’t need high concurrency were always written with treads, just because it was simpler.
similarly
Similarly, the FTP server in Harvest uses threads.
as a case in point
As a case in point, the cooperative task management technique described by Adya et al.