Annotation of libwww/Library/src/HTFilter.c, revision 2.6
2.1 frystyk 1: /*
2: ** BEFORE AND AFTER FILTERS
3: **
4: ** (c) COPYRIGHT MIT 1995.
5: ** Please first read the full copyright statement in the file COPYRIGH.
2.6 ! frystyk 6: ** @(#) $Id: HTFilter.c,v 2.5 1996/08/08 02:16:45 frystyk Exp $
2.1 frystyk 7: **
8: ** This module implrments a set of default filters that can be registerd
9: ** as BEFORE and AFTER filters to the Net manager
10: ** Authors
11: ** HFN Henrik Frystyk, frystyk@w.org
12: ** History
13: ** Jul 4, 96 Written
14: */
15:
16: /* Library include files */
17: #include "WWWLib.h"
18: #include "WWWCache.h"
19: #include "WWWRules.h"
20: #include "WWWHTTP.h"
21: #include "HTLog.h"
22: #include "HTAccess.h"
23: #include "HTFilter.h" /* Implemented here */
24:
25: /* ------------------------------------------------------------------------- */
26:
27: /*
28: ** Proxy and Gateway BEFORE filter
29: ** -------------------------------
30: ** Checks for registerd proxy servers or gateways and sees whether this
31: ** request should be redirected to a proxy or a gateway. Proxies have
32: ** higher priority than gateways so we look for them first!
33: ** For HTTP/1.0 and HTTP/1.1 we may only send a full URL (including the
34: ** host portion) to proxy servers. Therefore, we tell the Library whether
35: ** to use the full URL or the traditional HTTP one without the host part.
36: */
37: PUBLIC int HTProxyFilter (HTRequest * request, void * param, int status)
38: {
39: HTParentAnchor * anchor = HTRequest_anchor(request);
2.2 frystyk 40: char * addr = HTAnchor_physical(anchor);
2.1 frystyk 41: char * physical = NULL;
42: if ((physical = HTProxy_find(addr))) {
2.6 ! frystyk 43: HTRequest_setFullURI(request, YES); /* For now */
2.5 frystyk 44: HTRequest_setProxy(request, physical);
2.6 ! frystyk 45: #if 0
! 46: /* Don't paste the URLs together anymore */
2.1 frystyk 47: StrAllocCat(physical, addr);
2.5 frystyk 48: HTAnchor_setPhysical(anchor, physical);
2.6 ! frystyk 49: #endif
2.1 frystyk 50: } else if ((physical = HTGateway_find(addr))) {
51: /*
52: ** A gateway URL is crated by chopping off any leading "/" to make the
53: ** host into part of path
54: */
55: char * path =
56: HTParse(addr, "", PARSE_HOST + PARSE_PATH + PARSE_PUNCTUATION);
57: char * gatewayed = HTParse(path+1, physical, PARSE_ALL);
58: HTAnchor_setPhysical(anchor, gatewayed);
59: HT_FREE(path);
60: HT_FREE(gatewayed);
61: HTRequest_setFullURI(request, NO);
2.6 ! frystyk 62: HTRequest_deleteProxy(request);
2.1 frystyk 63: } else {
2.6 ! frystyk 64: HTRequest_setFullURI(request, NO); /* For now */
! 65: HTRequest_deleteProxy(request);
2.1 frystyk 66: }
67: return HT_OK;
68: }
69:
70: /*
71: ** Rule Translation BEFORE Filter
72: ** ------------------------------
73: ** If we have a set of rules loaded (see the Rule manager) then check
74: ** before each request whether how that should be translated. The trick
75: ** is that a parent anchor has a "address" which is the part from the URL
76: ** we used when we created the anchor. However, it also have a "physical
77: ** address" which is the place we are actually going to look for the
2.2 frystyk 78: ** resource. Hence this filter translates the physical address
79: ** (if any translations are found)
2.1 frystyk 80: */
81: PUBLIC int HTRuleFilter (HTRequest * request, void * param, int status)
82: {
83: HTList * list = HTRule_global();
84: HTParentAnchor * anchor = HTRequest_anchor(request);
2.2 frystyk 85: char * addr = HTAnchor_physical(anchor);
2.1 frystyk 86: char * physical = HTRule_translate(list, addr, NO);
87: if (!physical) {
88: HTRequest_addError(request, ERR_FATAL, NO, HTERR_FORBIDDEN,
89: NULL, 0, "HTRuleFilter");
90: return HT_ERROR;
91: }
92: HTAnchor_setPhysical(anchor, physical);
93: HT_FREE(physical);
94: return HT_OK;
95: }
96:
97: /*
98: ** Cache Validation BEFORE Filter
99: ** ------------------------------
100: ** Check the cache mode to see if we can use an already loaded version
101: ** of this document. If so and our copy is valid then we don't have
102: ** to go out and get it unless we are forced to
2.3 frystyk 103: ** We only check the cache in caseof a GET request. Otherwise, we go
104: ** directly to the source.
2.1 frystyk 105: */
106: PUBLIC int HTCacheFilter (HTRequest * request, void * param, int status)
107: {
108: HTParentAnchor * anchor = HTRequest_anchor(request);
109: HTReload mode = HTRequest_reloadMode(request);
2.3 frystyk 110: HTMethod method = HTRequest_method(request);
111:
112: /*
113: ** Check the method of the request
114: */
115: if (method != METHOD_GET) {
116: if (CACHE_TRACE) HTTrace("Cachefilter. We only check GET methods\n");
117: return HT_OK;
118: }
119:
2.1 frystyk 120: /*
121: ** If the mode if "Force Reload" then don't even bother to check the
122: ** cache - we flush everything we know about this document
123: */
124: if (mode == HT_FORCE_RELOAD) {
125: /*
126: ** Add the appropriate request headers. We use both the "pragma"
127: ** and the "cache-control" headers in order to be
128: ** backwards compatible with HTP/1.0
129: */
130: HTRequest_addGnHd(request, HT_G_PRAGMA_NO_CACHE);
131:
132: /* @@@ CACHE CONTROL @@@ */
133:
134: /*
135: ** We also flush the information in the anchor
136: */
137: HTAnchor_clearHeader(anchor);
138: return HT_OK;
139: }
140:
141: /*
142: ** Check the application provided memory cache. This is equivalent to a
143: ** history list and does not follow the same cache mechanisms as the
144: ** persistent cache
145: */
146: if (HTMemoryCache_check(request) == HT_LOADED)
147: return HT_LOADED;
148:
149: /*
150: ** Check the persistent cache manager. If we have a cache hit then
151: ** continue to see if the reload mode requires us to do a validation check.
152: ** This filter assumes that we can get the cached version through one of
153: ** our protocol modules (for example the file module)
154: */
155: {
156: char * addr = HTAnchor_address((HTAnchor *) anchor);
157: char * cache = HTCache_getReference(addr);
158: if (cache) {
159: if (mode != HT_CACHE_REFRESH) {
160: HTAnchor_setPhysical(anchor, cache);
161: HTAnchor_setCacheHit(anchor, YES);
162: } else {
163:
164: /* @@@ Do cache validation @@@ */
165:
166: }
167: }
168: HT_FREE(addr);
169: }
170: return HT_OK;
171: }
172:
173: /*
174: ** Error and Information AFTER filter
175: ** ----------------------------------
176: ** It checks the status code from a request and generates an
177: ** error/information message if required.
178: */
179: PUBLIC int HTInfoFilter (HTRequest * request, void * param, int status)
180: {
181: HTParentAnchor * anchor = HTRequest_anchor(request);
182: char * uri = HTAnchor_address((HTAnchor*) anchor);
183: switch (status) {
184: case HT_RETRY:
185: if (PROT_TRACE)
186: HTTrace("Load End.... NOT AVAILABLE, RETRY AT %ld\n",
187: HTRequest_retryTime(request));
188: break;
189:
190: case HT_ERROR:
191: {
192: /*
193: ** See if we have a function registered for outputting errors.
194: ** If so then call it and present the message to the user
195: */
196: HTAlertCallback *cbf = HTAlert_find(HT_A_MESSAGE);
197: if (cbf) (*cbf)(request, HT_A_MESSAGE, HT_MSG_NULL, NULL,
198: HTRequest_error(request), NULL);
199: if (PROT_TRACE)
200: HTTrace("Load End.... ERROR: Can't access `%s\'\n",
201: uri ? uri : "<UNKNOWN>");
2.3 frystyk 202: break;
203: }
204:
205: case HT_LOADED:
206: {
207: /*
208: ** Even though we have received a loaded status the thing we have
209: ** loaded successfully may in fact be an error message. We therefore
210: ** look at the error stack to see what to do.
211: */
212: HTAlertCallback *cbf = HTAlert_find(HT_A_MESSAGE);
213: if (cbf) (*cbf)(request, HT_A_MESSAGE, HT_MSG_NULL, NULL,
214: HTRequest_error(request), NULL);
215: if (PROT_TRACE) HTTrace("Load End.... OK: `%s\'\n", uri);
2.1 frystyk 216: break;
217: }
218:
219: default:
220: if (PROT_TRACE)
221: HTTrace("Load End.... Request ended with code %d\n", status);
222: break;
223: }
224:
225: HT_FREE(uri);
226: return HT_OK;
227: }
228:
229: /*
230: ** Redirection AFTER filter
231: ** ------------------------
232: ** The redirection handler only handles redirections
233: ** on the GET or HEAD method (or any other safe method)
234: */
235: PUBLIC int HTRedirectFilter (HTRequest * request, void * param, int status)
236: {
237: HTMethod method = HTRequest_method(request);
238: HTAnchor * new_anchor = HTRequest_redirection(request);
239:
240: /*
241: ** Only do redirect on GET and HEAD
242: */
243: if (!HTMethod_isSafe(method) || !new_anchor) {
2.4 frystyk 244: HTAlertCallback * prompt = HTAlert_find(HT_A_CONFIRM);
245: if (prompt) {
246: if ((*prompt)(request, HT_A_CONFIRM, HT_MSG_REDIRECTION,
247: NULL, NULL, NULL) != YES)
248: return HT_ERROR;
249: }
2.1 frystyk 250: }
251:
252: /*
253: ** Start new request with the redirect anchor found in the headers.
254: ** Note that we reuse the same request object which means that we must
255: ** keep this around until the redirected request has terminated. It also
256: ** allows us in an easy way to keep track of the number of redirections
257: ** so that we can detect endless loops.
258: */
2.4 frystyk 259: if (HTRequest_doRetry(request)) {
2.1 frystyk 260: HTLoadAnchor(new_anchor, request);
261: } else {
262: HTRequest_addError(request, ERR_FATAL, NO, HTERR_MAX_REDIRECT,
263: NULL, 0, "HTRedirectFilter");
264: }
265:
266: /*
267: ** By returning HT_ERROR we make sure that this is the last handler to be
268: ** called. We do this as we don't want any other filter to delete the
269: ** request object now when we have just started a new one ourselves
270: */
271: return HT_ERROR;
272: }
273:
274: /*
275: ** Client side authentication BEFORE filter
276: ** ----------------------------------------
277: ** The filter generates the credentials required to access a document
278: ** Getting the credentials may involve asking the user
279: */
280: PUBLIC int HTCredentialsFilter (HTRequest * request, void * param, int status)
281: {
282: /*
283: ** Ask the authentication module to call the right credentials generator
284: ** that understands this scheme
285: */
286: if (HTAA_beforeFilter(request, param, status) == HT_OK) {
287: if (PROT_TRACE) HTTrace("Credentials. verified\n");
288: return HT_OK;
289: } else {
290: HTRequest_addError(request, ERR_FATAL, NO, HTERR_UNAUTHORIZED,
291: NULL, 0, "HTCredentialsFilter");
292: return HT_ERROR;
293: }
294: }
295:
296: /*
297: ** Client side authentication AFTER filter
298: ** ---------------------------------------
299: ** The client side authentication filter uses the
300: ** user dialog messages registered in the HTAlert module.
301: ** By default these are the ones used by the line mode browser but you can
302: ** just register something else.
303: */
304: PUBLIC int HTAuthFilter (HTRequest * request, void * param, int status)
305: {
306: /*
307: ** Ask the authentication module to call the right challenge parser
308: ** that understands this scheme
309: */
310: if (HTAA_afterFilter(request, param, status) == HT_OK) {
311:
312: /*
313: ** Start request with new credentials. As with the redirection filter
314: ** we reuse the same request object which means that we must
315: ** keep this around until the redirected request has terminated
316: */
317: HTLoad(request, NO);
318:
319: /*
320: ** We return HT_ERROR to make sure that this is the last handler to be
321: ** called. We do this as we don't want any other filter to delete the
322: ** request object now when we have just started a new one ourselves
323: */
324: return HT_ERROR;
325: }
326: return HT_OK;
327: }
328:
329: /*
330: ** Request Loggin AFTER filter
331: ** ---------------------------
332: ** Default Logging filter using the log manager provided by HTLog.c
333: */
334: PUBLIC int HTLogFilter (HTRequest * request, void * param, int status)
335: {
336: if (request) {
337: if (HTLog_isOpen()) HTLog_add(request, status);
338: return HT_OK;
339: }
340: return HT_ERROR;
341: }
Webmaster