Re: [boost] Review: Boost.Threads

From: Lie-Quan Lee (llee1_at_[hidden])
Date: 2001-09-03 13:08:34

I think that the threading library exhibits sound design and implementation. 
I like that it is accepted and included into the boost libeary collection.

Here are some comments:

*) In source code, make member functions inline fo those only have few
   lines.  For example, mutex::do_lock(). Maybe combine some (if not
   all) src files to header as implicit inlining?

*) It is better to change the name tss to be more descriptive,
   something like thread_specific_storage is perfectly fine to me. But
   name tss conveys little.

*) In html documentation, it is better to reorganize to either
   Concepts
       Mutexes
       Lock
   Implementation
       mutex
       lock 
       thread
   ....
or 
   Mutex 
        concept
        class
   Lock
        concept
        class

Personally, I like the first way more. In that way, it separates
design and implemetation documentation. 

*) Some concepts are missing in the doc, for example, semaphore,
condition, thread, tss. I recalled that Jeremy made a few pages for
threading concepts (such as condition), it would be better to
incorprate them into the doc here.

*) In rationale.html, under "5.  Creation of a thread whose ownership is
passed to another object/method.", object manager deserves more
explanation. It looks like an object of thread_group type. But nothing
is mentioned in the rationale.html.

--
Lie-Quan Lee (AKA: Rich Lee)
Research Associate 
Computer Science Department
Indiana University