6 ashd - A Sane HTTP Daemon
11 This document describes the architecture and protocol of ashd
12 technically. If you want a brief overview, please see the homepage at
13 <http://www.dolda2000.com/~fredrik/ashd/>.
15 The basic premise of ashd is that of standard Unix philosophy; it
16 consists of a number of different programs, each specialized to one
17 precise task, passing HTTP requests around to each other in a manner
18 akin to standard Unix pipelines. This document describes the protocols
19 and conventions used between such programs that allows them to
25 All requests within ashd are created by *htparser*(1), which speaks
26 HTTP with its clients, translates the requests it receives into ashd
27 format, and passes them to a specified handler program. The handler
28 program may choose to respond to the request itself, or pass it on to
29 another handler for further processing.
31 A request in ashd format consists of 4 structural parts:
33 HTTP method, URL and version::
35 The HTTP header line information, exactly as specified by the
36 client. That is, any escape sequences in the URL are passed
37 around in non-processed form, since each program needs to
38 handle escape processing in its own way.
42 The 'rest string' (sometimes referred to as the 'point', in
43 deference to Emacs parlance) is the part of the URL which
44 remains to be processed. Each handler program is free to
45 modify the rest string (usually, but not necessarily, by
46 removing leading parts of it) before passing the request on to
47 another handler. When *htparser*(1) initially constructs a
48 request, it forms the rest string from the URL by stripping
49 off the initial slash and the query parameters. In other
50 words, a request to `/a/b/c?d=e` will be given the initial
55 The HTTP headers are parsed and passed along with the request
56 as they are, but *htparser*(1) itself, and some handler
57 programs, add some more headers, prefixed with `X-Ash-`,
58 and to safeguard any potentially sensitive such headers from
59 forgery, *htparser*(1) strips any headers with that prefix
60 from the incoming request.
64 Along with the request information, a socket for responding is
65 passed. A handler program that wishes to actually respond to a
66 request needs only output a well-formed HTTP response on this
67 socket and then close it. The details are covered below, but
68 note that the socket is connected to *htparser*(1) rather than
69 the client itself, and that *htparser* will do any transfer
70 encoding that may be required for HTTP keep-alive. The
71 response socket is also used for reading the request-body, if
72 the client provides one.
77 Handler programs are started either by *htparser*(1) itself, or in
78 turn by other handler programs, and certain conventions are observed
81 There are two basic types of handler programs, persistent and
82 transient, which determines the convention used in starting them. A
83 persistent program will continue running indefinitely, and handle any
84 amount of requests during its lifetime, while a transient program will
85 handle one request only and then exit. The convention of transient
86 programs was created mainly for convenience, since it is easier to
87 write such programs. The *htparser*(1) program will only start a
88 persistent program as the root handler.
90 A persistent handler program, when started, is passed a Unix socket of
91 SEQPACKET type on standard input. Its parent program will then pass
92 one datagram per request on that socket, containing the above listed
93 parts of the request using the datagram format described below. By
94 convention, the handler program should exit normally if it receives
95 end-of-file on the socket.
97 A transient program, when started, has the response socket connected
98 to its standard input and output (standard error is inherited from the
99 parent process). It may be provided arbitrary arguments as supplied by
100 the program starting it, but the last three arguments are the HTTP
101 method, the raw URL and the rest string, in that order. The HTTP
102 headers are converted into environment variables by turning them into
103 uppercase, replacing any dashs with underscores, and prefixing them
104 with `REQ_`. For example, the HTTP `Host` header will be passed as
105 `REQ_HOST`. The HTTP protocol version is passed in an environment
106 variable called `HTTP_VERSION`. It is passed in full; i.e. as
107 `HTTP/1.1`, rather than just `1.1`.
109 The response socket, as mentioned above, is also used for reading the
110 request-body if the client provides one. For such purposes,
111 *htparser*(1) ensures that the reader sees end-of-file at the end of
112 the request-body, so that the reader (unlike in, for example, CGI)
113 does not have to worry about the Content-Length header and counting
116 To respond, the handler program needs to write an ordinary HTTP
117 response to the response socket. That is, one line containing the HTTP
118 version, status code and status text, followed by any number of lines
119 with headers, followed by an empty line, followed by the
120 response-body. Basic familiarity with HTTP should relieve this
121 document of detailing the exact format of such a response, but the
122 following points are noteworthy:
124 * The HTTP version is actually ignored; it must simply be there for
127 * In the header, Unix line endings are accepted; *htparser*(1) will
128 still use CRLF line endings when passing the response to the
131 * The response socket should be closed when the entire body has been
132 written. *htparser*(1) itself will take care of anything needed for
133 HTTP keep-alive, such as chunking. It is recommended, however, that
134 the handler program provides the Content-Length header if it can be
135 calculated in advance, since *htparser*(1) will not need to add
136 chunking in such cases.
138 * *htparser*(1) will not provide an error message to the client in
139 case the response socket is closed before a complete response has
140 been written to it, so a handler program should always provide an
141 error message by itself if a request cannot be handled for some
147 The datagram format used for persistent handler programs is simply a
148 sequence of NUL-terminated strings. The datagram starts with the HTTP
149 method, the URL, the HTTP version and the rest string, in that
150 order. They are followed by an arbitrary number of string pairs, one
151 for each header; the first string in a pair being the header name, and
152 the second being the value. The headers are terminated by one instance
155 Along with the datagram, the response socket is passed using the
156 SCM_RIGHTS ancillary message for Unix sockets. See *unix*(7),
157 *recvmsg*(2) and *sendmsg*(2) for more information. Each datagram will
158 have exactly one associated socket passed with it.
162 Fredrik Tolf <fredrik@dolda2000.com>