| 1 |
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd"> |
| 2 |
<html><head> |
| 3 |
<title>liboop: Why?</title> |
| 4 |
<link rel="stylesheet" type="text/css" href="style.css"> |
| 5 |
</head><body> |
| 6 |
|
| 7 |
<h2>Why use liboop?</h2> |
| 8 |
|
| 9 |
<h4>The problem.</h4> |
| 10 |
|
| 11 |
Developers often wish to write applications which serve as a mediator between |
| 12 |
several logical interfaces simultaneously; in fact, most applications work |
| 13 |
this way. For example, a browser application might wish to maintain a user |
| 14 |
interface while also managing a network connection and occasionally exchanging |
| 15 |
data with the local filesystem. A server application might be communicating |
| 16 |
with several clients at once while also occasionally receiving a signal from |
| 17 |
the administrator directing it to reload its configuration. A multiplayer game |
| 18 |
might want to maintain several active user interfaces at once. |
| 19 |
<p> |
| 20 |
Furthermore, each of these interfaces may be quite complex, sufficiently so to |
| 21 |
merit shared code modules which specialize in managing the interface. |
| 22 |
Widget sets deal with the details of the X protocol and graphical user |
| 23 |
interface management; "curses" deals with the arcana of character-based |
| 24 |
terminals; WWW libraries offer high-level access to whole families of Internet |
| 25 |
transfer protocols; standard I/O and database routines manage filesystem data. |
| 26 |
<p> |
| 27 |
However, the existing techniques available for multiplexing interface code are |
| 28 |
very poor. Most of these libraries work in "blocking" fashion; once |
| 29 |
instructed to complete a task (such as downloading a file, or presenting a |
| 30 |
dialog to the user), they do not return until the task is complete (or failed), |
| 31 |
even though this may mean waiting an arbitrary amount of time for some external |
| 32 |
agent (such as the user or the network) to respond. Some of the better systems |
| 33 |
are able to manage several concurrent tasks internally, but cannot work with |
| 34 |
other components. |
| 35 |
<p> |
| 36 |
Developers are thus left with several unpalatable choices: |
| 37 |
<ol> |
| 38 |
<li>Accept "blocking" operation. User interfaces stop functioning while the |
| 39 |
application waits for the network; one network client's access is stalled |
| 40 |
while another client performs a transaction. As more data moves from local |
| 41 |
storage (where access is fast enough that blocking is acceptable) to |
| 42 |
delay-prone networked media, this is becoming less and less acceptable. |
| 43 |
<li>Use multiple threads for concurrency. While this is a good solution for |
| 44 |
some problems, developers who choose this route must struggle with relatively |
| 45 |
immature and unportable threading models, and deal with the many libraries |
| 46 |
which are not thread-safe; furthermore, threaded programming requires |
| 47 |
thought-intensive and error-prone synchronization. |
| 48 |
<li>Use multiple processes ("forking") for concurrency. This can also work, |
| 49 |
but requires all communication between modules to use some form of |
| 50 |
inter-process communication, which increases complexity and decreases |
| 51 |
performance. Forking itself is a slow operation, leading to complex |
| 52 |
"pre-forking" schemes for better performance. Worst of all, each process |
| 53 |
must somehow multiplex IPC from other processes with whatever I/O task it had |
| 54 |
to accomplish in the first place; this brings back the very problem forking |
| 55 |
was designed to address. |
| 56 |
<li>Attempt to multiplex each library's I/O operations directly in a master |
| 57 |
"select loop". This requires the developer to understand intimately the |
| 58 |
exact details of each library's I/O interactions, thus breaking modularity, |
| 59 |
fostering unhealthy dependency and leading to a single central snarl through |
| 60 |
which all I/O must pass. |
| 61 |
</ol> |
| 62 |
The paucity of options is reflected in the quality of applications. How many |
| 63 |
programs hang unpleasantly while performing simple network operations like |
| 64 |
hostname resolution? How many user interfaces are unnecessarily "modal"? |
| 65 |
How many simple servers fork for no good reason? How many network applications |
| 66 |
simply don't exist because it's so difficult to write them? |
| 67 |
|
| 68 |
<h4>The solution.</h4> |
| 69 |
|
| 70 |
Liboop offers a single, simple, central event loop. Modules wishing to perform |
| 71 |
I/O without blocking request <em>callbacks</em> from the central <em>event |
| 72 |
source</em>. These callbacks may be tied to file-descriptor activity, the |
| 73 |
system time, or process signals. Liboop is responsible for invoking these |
| 74 |
callbacks as appropriate. |
| 75 |
<p> |
| 76 |
With this system, each module "owns" its own I/O; it can perform arbitrarily |
| 77 |
complex operations without blocking anything else in the program. But since |
| 78 |
callbacks are executed purely sequentially, there is no complex concurrent code |
| 79 |
to manage. From the application developer's point of view, working with liboop |
| 80 |
is very simple; the developer simply makes calls to libraries which work their |
| 81 |
magic and call the application back when they finish. Applications can easily |
| 82 |
manage an arbitrary amount of multiplexed I/O operations using as many |
| 83 |
interface libraries as they like without blocking. |
| 84 |
<p> |
| 85 |
To work with this system, libraries and applications must be liboop-aware. |
| 86 |
Development with legacy code uses <em>adapters</em> which translate the I/O |
| 87 |
model of an application or library into liboop's model. This does require |
| 88 |
knowledge of the code's I/O structure, but can at least keep the modules in |
| 89 |
an application independent of each other. |
| 90 |
<p> |
| 91 |
For more about liboop, see the <a href="how.html">documentation</a>. |
| 92 |
|
| 93 |
<h4>Q&A</h4> |
| 94 |
|
| 95 |
<dl> |
| 96 |
<dt><em>Why don't you just use (favorite widget set), which lets you register |
| 97 |
callbacks on file descriptors and all that good stuff?</em> |
| 98 |
<dd>Because not everyone might want to be tied to that widget set. In |
| 99 |
particular, the developer of a general-purpose I/O library would want to |
| 100 |
allow everyone to use it, without requiring a particular widget set. |
| 101 |
Liboop lets the library developer write to a standard interface, |
| 102 |
which can then be used with most widget sets and other event loops.<p> |
| 103 |
|
| 104 |
<a name="glib"></a> |
| 105 |
<dt><em>Doesn't GLib's <a |
| 106 |
href="http://developer.gnome.org/doc/API/glib/glib-the-main-event-loop.html">Main |
| 107 |
Event Loop</a> do all this, and more?</em> |
| 108 |
<dd>Not quite. GLib is a fine implementation of an event loop (with |
| 109 |
bells and whistles) that supports some extensibility (such as the ability to |
| 110 |
add extra sources). However, I'm doubtful that it extends far enough that |
| 111 |
it could run on top of someone else's event loop (such as the Tk event loop). |
| 112 |
Furthermore, the GLib event loop doesn't manage signals; synchronous handling |
| 113 |
of asynchronous signals is very difficult to do properly and safely in most |
| 114 |
existing systems (without kludges like polling). |
| 115 |
|
| 116 |
<p>In any case, we do have a |
| 117 |
<a href="oop_glib.html">GLib source adapter</a> so you can use the GLib event loop |
| 118 |
with the liboop interface.</p> |
| 119 |
|
| 120 |
<dt><em>How does liboop compare to Niels Provos' <a |
| 121 |
href="http://www.monkey.org/~provos/libevent/">libevent</a>?</em> |
| 122 |
<dd>Like GLib, libevent is a concrete implementation of an event loop, not |
| 123 |
an abstract interface for many event loops; also like GLib, libevent does not |
| 124 |
manage signals. Libevent is smaller and simpler than either liboop or Glib. |
| 125 |
While liboop and GLib are both licensed under the |
| 126 |
<a href="http://www.fsf.org/copyleft/lesser.html">Lesser GPL</a>, libevent |
| 127 |
appears to be licensed under the original BSD license, including the |
| 128 |
advertising clause. Note that the advertising clause renders libevent |
| 129 |
incompatible with GPL software! |
| 130 |
|
| 131 |
<p>It is entirely possible to imagine a libevent source adapter for liboop. |
| 132 |
If anyone is interested in such an adapter, please contact me.</p> |
| 133 |
|
| 134 |
</dl> |
| 135 |
|
| 136 |
<hr><a href="index.html">liboop home</a></body></html> |