1 $Id: config-design 4805 2001-06-21 10:52:27Z rra $
3 This file is documentation of the design principles that went into INN's
4 configuration file syntax, and some rationale for why those principles
7 1. All configuration files used by INN should have the same syntax.
8 This was the root reason why the project was taken on in the first
9 place; INN developed a proliferation of configuration files, all of
10 which had a slightly (or greatly) different syntax, forcing the
11 administrator to learn several different syntaxes and resulting in a
12 proliferation of parsers, all with their own little quirks.
14 2. Adding a new configuration file or a new set of configuration options
15 should not require writing a single line of code for syntax parsing.
16 Code that analyzes the semantics of the configuration will of course
17 be necessary, but absolutely no additional code to read files, parse
18 files, build configuration trees, or the like should be required.
19 Ideally, INN should have a single configuration parser that
22 3. The syntax should look basically like the syntax of readers.conf,
23 incoming.conf, and innfeed.conf in INN 2.3. After extensive
24 discussion on the inn-workers mailing list, this seemed to be the
25 most generally popular syntax of the ones already used in INN, and
26 inventing a completely new syntax didn't appear likely to have gains
27 outweighing the effort involved. This syntax seemed sufficiently
28 general to represent all of the configuration information that INN
31 4. The parsing layer should *not* attempt to do semantic analysis of the
32 configuration; it should concern itself solely with syntax (or very
33 low-level semantics that are standard across all conceivable INN
34 configuration files). In particular, the parsing layer should not
35 know what parameters are valid, what groups are permitted, what types
36 the values for parameters should have, or what default values
39 This principle requires some additional explanation, since it is very
40 tempting to not do things this way. However, the more semantic
41 information the parser is aware of, the less general the parser is,
42 and it's very easy to paint oneself into a corner. In particular,
43 it's *not* a valid assumption that all clients of the parsing code
44 will want to reduce the configuration to a bunch of structs; this
45 happens to be true for most clients of inn.conf, for example, but
46 inndstart doesn't want the code needed to reduce everything to a
47 struct and set default values to necessarily be executed in a
48 security-critical context.
50 Additionally, making the parser know more semantic information either
51 complicates (significantly) the parser interface or means that the
52 parser has to be modified when the semantics change. The latter is
53 not acceptable, and the parser interface should be as straightforward
54 as possible (to encourage all parts of INN to use it).
56 5. The result of a parse of the configuration file may be represented as
57 a tree of dictionaries, where each dictionary corresponds to a group
58 and each key corresponds to a parameter setting. (Note that this does
59 not assume that the underlying data structure is a hash table, just
60 that it has dictionary semantics, namely a collection of key/value
61 pairs with the keys presumed unique.)
63 6. Parameter values inherit via group nesting. In other words, if a
64 group is nested inside another group, all parameters defined in the
65 enclosing group are inherited by the nested group unless they're
66 explicitly overriden within the nested group. (This point and point
67 5 are to some degree just corollaries of point 3.)
69 7. The parsing library must permit writing as well as reading. It must
70 be possible for a program to read in a configuration file, modify
71 parameters, add and delete groups, and otherwise change the
72 configuration, and then write back out to disk a configuration file
73 that preserves those changes and still remains as faithful to the
74 original (possibly human-written) configuration file as possible.
75 (Ideally, this would extend to preserving comments, but that may be
76 too difficult to do and therefore isn't required.)
78 8. The parser must not limit the configuration arbitrarily. In
79 particular, unlimited length strings (within available memory) must
80 be supported for string values, and if allowable line length is
81 limited, line continuation must be supported everywhere that there's
82 any reasonable expectation that it might be necessary. One common
83 configuration parameter is a list of hosts or host wildmats that can
84 be almost arbitrarily long, and the syntax and parser must support
87 9. The parser should be reasonably efficient, enough so as to not cause
88 an annoying wait for command-line tools like sm and grephistory to
89 start. In general, though, efficiency in either time or memory is
90 not as high of a priority as readable, straightforward code; it's
91 safe to assume that configuration parsing is only done on startup and
92 at rare intervals and is not on any critical speed paths.
94 10. Error reporting is a must. It must be possible to clearly report
95 errors in the configuration files, including at minimum the file name
96 and line number where the error occurred.
98 11. The configuration parser should not trust its input, syntax-wise. It
99 must not segfault, infinitely loop, or otherwise explode on malformed
100 or broken input. And, as a related point, it's better to be
101 aggressively picky about syntax than to be lax and attempt to accept
102 minor violations. The intended configuration syntax is simple and
103 unambiguous, so it should be unnecessary to accept violations.
105 12. It must be possible to do comprehensive semantic checks of a
106 configuration file, including verifying that all provided parameters
107 are known ones, all parameter values have the correct type, group
108 types that are not expected to be repeated are not, and only expected
109 group types are used. This must *not* be done by the parser, but the
110 parser must provide sufficient hooks that the client program can do
113 13. The parser must be re-entrant and thread-safe.
115 14. The grammar shouldn't require any lookahead to parse. This is in
116 order to keep the parser extremely simple and therefore maintainable.
117 (It's worth noting that this design principle leads to the
118 requirement that parameter keys end in a colon; the presence of the
119 colon allows parameter keys to be distinguished from other syntactic
120 elements allowed in the same scope, like the beginning of a nested