Annotation of libwww/Library/src/HTTP.c, revision 1.18
1.1 timbl 1: /* HyperText Tranfer Protocol - Client implementation HTTP.c
2: ** ==========================
1.2 timbl 3: **
4: ** Bugs:
5: ** Not implemented:
6: ** Forward
7: ** Redirection
8: ** Error handling
1.1 timbl 9: */
10:
11: /* Module parameters:
12: ** -----------------
13: **
14: ** These may be undefined and redefined by syspec.h
15: */
1.2 timbl 16:
1.12 timbl 17: /* MOSAIC_HACK2 is a kludge to guess the file type of trabsferred
18: ** file from the URL. It is STRICTLY illegal to do this!
19: */
20:
1.2 timbl 21: /* Implements:
22: */
23: #include "HTTP.h"
24:
25: #define HTTP_VERSION "HTTP/1.0"
26: #define HTTP2 /* Version is greater than 0.9 */
27:
28: #define INIT_LINE_SIZE 1024 /* Start with line buffer this big */
29: #define LINE_EXTEND_THRESH 256 /* Minimum read size */
30: #define VERSION_LENGTH 20 /* for returned protocol version */
31:
32: /* Uses:
33: */
1.1 timbl 34: #include "HTParse.h"
35: #include "HTUtils.h"
36: #include "tcp.h"
37: #include "HTTCP.h"
38: #include "HTFormat.h"
1.2 timbl 39: #include <ctype.h>
40: #include "HTAlert.h"
41: #include "HTMIME.h"
1.5 timbl 42: #include "HTML.h" /* SCW */
43: #include "HTInit.h" /* SCW */
1.14 luotonen 44: #include "HTAABrow.h" /* Access Authorization */
1.1 timbl 45:
1.2 timbl 46: struct _HTStream {
47: HTStreamClass * isa; /* all we need to know */
48: };
49:
50:
1.6 timbl 51: extern char * HTAppName; /* Application name: please supply */
52: extern char * HTAppVersion; /* Application version: please supply */
53:
1.1 timbl 54: /* Load Document from HTTP Server HTLoadHTTP()
55: ** ==============================
56: **
57: ** Given a hypertext address, this routine loads a document.
58: **
59: **
60: ** On entry,
61: ** arg is the hypertext reference of the article to be loaded.
62: **
63: ** On exit,
64: ** returns >=0 If no error, a good socket number
65: ** <0 Error.
66: **
67: ** The socket must be closed by the caller after the document has been
68: ** read.
69: **
70: */
1.17 timbl 71: PUBLIC int HTLoadHTTP ARGS2 (
1.2 timbl 72: CONST char *, arg,
1.17 timbl 73: HTRequest *, request)
1.1 timbl 74: {
75: int s; /* Socket number for returned data */
76: int status; /* tcp return */
1.10 timbl 77: char crlf[3]; /* A CR LF equivalent string */
1.3 timbl 78: HTStream * target = NULL; /* Unconverted data */
1.15 luotonen 79: char *auth = NULL; /* Authorization information */
1.3 timbl 80:
1.2 timbl 81: CONST char* gate = 0; /* disable this feature */
1.1 timbl 82: SockA soc_address; /* Binary network address */
83: SockA * sin = &soc_address;
1.2 timbl 84: BOOL extensions = YES; /* Assume good HTTP server */
1.17 timbl 85:
1.1 timbl 86: if (!arg) return -3; /* Bad if no name sepcified */
87: if (!*arg) return -2; /* Bad if name had zero length */
88:
89: /* Set up defaults:
90: */
91: #ifdef DECNET
1.2 timbl 92: sin->sdn_family = AF_DECnet; /* Family = DECnet, host order */
93: sin->sdn_objnum = DNP_OBJ; /* Default: http object number */
1.1 timbl 94: #else /* Internet */
1.2 timbl 95: sin->sin_family = AF_INET; /* Family = internet, host order */
96: sin->sin_port = htons(TCP_PORT); /* Default: http port */
1.1 timbl 97: #endif
98:
1.10 timbl 99: sprintf(crlf, "%c%c", CR, LF); /* To be corect on Mac, VM, etc */
100:
1.1 timbl 101: if (TRACE) {
102: if (gate) fprintf(stderr,
103: "HTTPAccess: Using gateway %s for %s\n", gate, arg);
104: else fprintf(stderr, "HTTPAccess: Direct access for %s\n", arg);
105: }
106:
107: /* Get node name and optional port number:
108: */
109: {
110: char *p1 = HTParse(gate ? gate : arg, "", PARSE_HOST);
111: int status = HTParseInet(sin, p1); /* TBL 920622 */
112: free(p1);
113: if (status) return status; /* No such host for example */
114: }
115:
1.2 timbl 116: retry:
1.15 luotonen 117:
118: /*
119: ** Compose authorization information (this was moved here
120: ** from after the making of the connection so that the connection
121: ** wouldn't have to wait while prompting username and password
122: ** from the user). -- AL 13.10.93
123: */
124: #ifdef ACCESS_AUTH
125: #define FREE(x) if (x) {free(x); x=NULL;}
126: {
127: char *docname;
128: char *hostname;
129: char *colon;
130: int portnumber;
131:
132: docname = HTParse(arg, "", PARSE_PATH);
133: hostname = HTParse((gate ? gate : arg), "", PARSE_HOST);
134: if (hostname &&
135: NULL != (colon = strchr(hostname, ':'))) {
1.16 duns 136: *(colon++) = '0円'; /* Chop off port number */
1.15 luotonen 137: portnumber = atoi(colon);
138: }
139: else portnumber = 80;
140:
141: auth = HTAA_composeAuth(hostname, portnumber, docname);
142:
143: if (TRACE) {
144: if (auth)
145: fprintf(stderr, "HTTP: Sending authorization: %s\n", auth);
146: else
147: fprintf(stderr, "HTTP: Not sending authorization (yet)\n");
148: }
149: FREE(hostname);
150: FREE(docname);
151: }
152: #endif /* ACCESS_AUTH */
1.1 timbl 153:
1.10 timbl 154: /* Now, let's get a socket set up from the server for the data:
1.1 timbl 155: */
156: #ifdef DECNET
157: s = socket(AF_DECnet, SOCK_STREAM, 0);
158: #else
159: s = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP);
160: #endif
161: status = connect(s, (struct sockaddr*)&soc_address, sizeof(soc_address));
162: if (status < 0) {
163: if (TRACE) fprintf(stderr,
164: "HTTP: Unable to connect to remote host for `%s' (errno = %d).\n", arg, errno);
1.17 timbl 165:
1.1 timbl 166: return HTInetStatus("connect");
167: }
168:
169: if (TRACE) fprintf(stderr, "HTTP connected, socket %d\n", s);
170:
1.17 timbl 171:
172: /* Compose and send command
173: ** ------------------------
174: */
175: {
176: char *command; /* The whole command */
177:
1.1 timbl 178: /* Ask that node for the document,
179: ** omitting the host name & anchor if not gatewayed.
180: */
1.17 timbl 181: if (gate) {
182: command = malloc(4 + strlen(arg)+ 2 + 31);
183: if (command == NULL) outofmem(__FILE__, "HTLoadHTTP");
184: strcpy(command, "GET ");
185: strcat(command, arg);
186: } else { /* not gatewayed */
187: char * p1 = HTParse(arg, "", PARSE_PATH|PARSE_PUNCTUATION);
188: command = malloc(4 + strlen(p1)+ 2 + 31);
189: if (command == NULL) outofmem(__FILE__, "HTLoadHTTP");
190: strcpy(command, "GET ");
191: strcat(command, p1);
192: free(p1);
193: }
1.2 timbl 194: #ifdef HTTP2
1.17 timbl 195: if (extensions) {
196: strcat(command, " ");
197: strcat(command, HTTP_VERSION);
198: }
1.2 timbl 199: #endif
1.17 timbl 200:
201: strcat(command, crlf); /* CR LF, as in rfc 977 */
202:
203: if (extensions) {
204:
205: int n;
206: int i;
207: HTAtom * present = WWW_PRESENT;
208: char line[256]; /*@@@@ */
209:
210: /* if (!request->conversions) HTFormatInit(request->conversions); */
211: n = HTList_count(request->conversions);
212:
213: for(i=0; i<n; i++) {
214: HTPresentation * pres =
215: HTList_objectAt(request->conversions, i);
216: if (pres->rep_out == present) {
217: if (pres->quality != 1.0) {
218: sprintf(line, "Accept: %s q=%.3f%c%c",
219: HTAtom_name(pres->rep), pres->quality, CR, LF);
220: } else {
221: sprintf(line, "Accept: %s%c%c",
222: HTAtom_name(pres->rep), CR, LF);
223: }
224: StrAllocCat(command, line);
225:
226: }
1.2 timbl 227: }
1.17 timbl 228:
229: sprintf(line, "User-Agent: %s/%s libwww/%s%c%c",
230: HTAppName ? HTAppName : "unknown",
231: HTAppVersion ? HTAppVersion : "0.0",
232: HTLibraryVersion, CR, LF);
233: StrAllocCat(command, line);
234:
1.14 luotonen 235: #ifdef ACCESS_AUTH
1.17 timbl 236: if (auth != NULL) {
237: sprintf(line, "%s%c%c", auth, CR, LF);
238: StrAllocCat(command, line);
239: }
240: #endif /* ACCESS_AUTH */
1.14 luotonen 241: }
1.17 timbl 242:
243: StrAllocCat(command, crlf); /* Blank line means "end" */
244:
245: if (TRACE) fprintf(stderr, "HTTP Tx: %s\n", command);
246:
247: /* Translate into ASCII if necessary
248: */
1.4 timbl 249: #ifdef NOT_ASCII
1.17 timbl 250: {
251: char * p;
252: for(p = command; *p; p++) {
253: *p = TOASCII(*p);
254: }
1.1 timbl 255: }
1.3 timbl 256: #endif
1.17 timbl 257:
258: status = NETWRITE(s, command, (int)strlen(command));
259: free(command);
260: if (status<0) {
261: if (TRACE) fprintf(stderr,
262: "HTTPAccess: Unable to send command.\n");
1.1 timbl 263: return HTInetStatus("send");
1.17 timbl 264: }
265: } /* compose and send command */
266:
1.2 timbl 267:
1.17 timbl 268: /* Read the response
269: ** -----------------
1.11 timbl 270: **
271: ** HTTP0 servers must return ASCII style text, though it can in
272: ** principle be just text without any markup at all.
273: ** Full HTTP servers must return a response
274: ** line and RFC822 style header. The response must therefore in
275: ** either case have a CRLF somewhere soon.
276: **
277: ** This is the theory. In practice, there are (1993) unfortunately
278: ** many binary documents just served up with HTTP0.9. This
279: ** means we have to preserve the binary buffer (on the assumption that
280: ** conversion from ASCII may lose information) in case it turns
281: ** out that we want the binary original.
1.2 timbl 282: */
1.3 timbl 283:
1.17 timbl 284: { /* read response */
285:
286: char * eol = 0; /* End of line if found */
287: char * start_of_data; /* Start of body of reply */
288: int length; /* Number of valid bytes in buffer */
289: char * text_buffer = NULL;
290: char * binary_buffer = NULL;
291: HTFormat format_in; /* Format arriving in the message */
292:
293: { /* local variablees for loop*/
1.2 timbl 294:
295: /* Get numeric status etc */
296:
1.17 timbl 297: BOOL end_of_file = NO;
298: HTAtom * encoding = HTAtom_for("7bit");
299: int buffer_length = INIT_LINE_SIZE; /* Why not? */
300:
301: binary_buffer = (char *) malloc(buffer_length * sizeof(char));
302: if (!binary_buffer) outofmem(__FILE__, "HTLoadHTTP");
303: text_buffer = (char *) malloc(buffer_length * sizeof(char));
304: if (!text_buffer) outofmem(__FILE__, "HTLoadHTTP");
305: length = 0;
1.2 timbl 306:
1.17 timbl 307: do { /* Loop to read in the first line */
308:
309: /* Extend line buffer if necessary for those crazy WAIS URLs ;-) */
310:
311: if (buffer_length - length < LINE_EXTEND_THRESH) {
312: buffer_length = buffer_length + buffer_length;
313: binary_buffer = (char *) realloc(
314: binary_buffer, buffer_length * sizeof(char));
315: if (!binary_buffer) outofmem(__FILE__, "HTLoadHTTP");
316: text_buffer = (char *) realloc(
317: text_buffer, buffer_length * sizeof(char));
318: if (!text_buffer) outofmem(__FILE__, "HTLoadHTTP");
319: }
320: status = NETREAD(s, binary_buffer + length,
321: buffer_length - length -1);
322: if (status < 0) {
323: HTAlert("Unexpected network read error on response");
324: NETCLOSE(s);
325: return status;
326: }
327:
328: if (TRACE) fprintf(stderr, "HTTP: read returned %d bytes.\n",
329: status);
330:
331: if (status == 0) {
332: end_of_file = YES;
333: break;
334: }
335: binary_buffer[length+status] = 0;
336:
337:
1.11 timbl 338: /* Make an ASCII *copy* of the buffer
339: */
1.2 timbl 340: #ifdef NOT_ASCII
1.17 timbl 341: if (TRACE) fprintf(stderr,
342: "Local codes CR=%d, LF=%d\n", CR, LF);
1.11 timbl 343: #endif
1.17 timbl 344: {
345: char * p;
346: char * q;
347: for(p = binary_buffer+length, q=text_buffer+length;
348: *p; p++, q++) {
349: *q = FROMASCII(*p);
350: }
351:
352: *q++ = 0;
1.11 timbl 353: }
1.17 timbl 354:
1.11 timbl 355: /* Kludge to trap binary responses from illegal HTTP0.9 servers.
356: ** First time we have enough, look at the stub in ASCII
357: ** and get out of here if it doesn't look right.
358: **
359: ** We also check for characters above 128 in the first few bytes, and
360: ** if we find them we forget the html default.
361: **
362: ** Bugs: A HTTP0.9 server returning a document starting "HTTP/"
363: ** will be taken as a HTTP 1.0 server. Failure.
364: ** An HTTP 0.9 server returning a binary document with
365: ** characters < 128 will be read as ASCII.
366: */
367: #define STUB_LENGTH 20
1.17 timbl 368: if (length < STUB_LENGTH && length+status >= STUB_LENGTH) {
369: if(strncmp("HTTP/", text_buffer, 5)!=0) {
370: char *p;
371: start_of_data = text_buffer; /* reparse whole reply */
372: for(p=binary_buffer; p <binary_buffer+STUB_LENGTH;p++) {
373: if (((int)*p)&128) {
374: format_in = HTAtom_for("www/unknown");
375: length = length + status;
376: goto copy; /* out of while loop */
377: }
1.11 timbl 378: }
379: }
1.2 timbl 380: }
1.17 timbl 381: /* end kludge */
382:
383:
384: eol = strchr(text_buffer + length, 10);
385: if (eol) {
386: *eol = 0; /* Terminate the line */
387: if (eol[-1] == CR) eol[-1] = 0; /* Chop trailing CR */
388: /* = corrected to == -- AL */
389: }
390:
391: length = length + status;
392:
393: } while (!eol && !end_of_file); /* No LF */
394:
395: } /* Scope of loop variables */
1.2 timbl 396:
1.7 timbl 397:
398: /* We now have a terminated unfolded line. Parse it.
399: ** -------------------------------------------------
1.2 timbl 400: */
401:
1.17 timbl 402: if (TRACE)fprintf(stderr, "HTTP: Rx: %.70s\n", text_buffer);
403:
404: {
405: int fields;
406: char server_version [VERSION_LENGTH+1];
407: int server_status;
408:
409:
410: #ifdef OLD_CODE /* old buggy servers should not exist now tbl9311 */
1.2 timbl 411: /* Kludge to work with old buggy servers. They can't handle the third word
412: ** so we try again without it.
413: */
1.17 timbl 414: if (extensions &&
415: 0==strcmp(text_buffer, /* Old buggy server? */
416: "Document address invalid or access not authorised")) {
417: extensions = NO;
418: if (binary_buffer) free(binary_buffer);
419: if (text_buffer) free(text_buffer);
420: if (TRACE) fprintf(stderr,
421: "HTTP: close socket %d to retry with HTTP0\n", s);
422: NETCLOSE(s);
423: goto retry; /* @@@@@@@@@@ */
424: }
425: #endif
426:
427: fields = sscanf(text_buffer, "%20s%d",
428: server_version,
429: &server_status);
430:
431: if (fields < 2 ||
432: strncmp(server_version, "HTTP/", 5)!=0) { /* HTTP0 reply */
433: format_in = WWW_HTML;
434: start_of_data = text_buffer; /* reread whole reply */
435: if (eol) *eol = '\n'; /* Reconstitute buffer */
436:
437: } else { /* Full HTTP reply */
1.2 timbl 438:
1.17 timbl 439: /* Decode full HTTP response */
1.2 timbl 440:
1.17 timbl 441: format_in = HTAtom_for("www/mime");
442: start_of_data = eol ? eol + 1 : text_buffer + length;
443:
444: switch (server_status / 100) {
1.3 timbl 445:
1.17 timbl 446: default: /* bad number */
447: HTAlert("Unknown status reply from server!");
448: break;
449:
450: case 3: /* Various forms of redirection */
451: HTAlert(
452: "Redirection response from server is not handled by this client");
453: break;
454:
455: case 4: /* Access Authorization problem */
1.14 luotonen 456: #ifdef ACCESS_AUTH
1.17 timbl 457: switch (server_status) {
458: case 401:
459: length -= start_of_data - text_buffer;
460: if (HTAA_shouldRetryWithAuth(start_of_data, length, s)) {
461: /* Clean up before retrying */
462: if (binary_buffer) free(binary_buffer);
463: if (text_buffer) free(text_buffer);
464: if (TRACE)
465: fprintf(stderr, "%s %d %s\n",
466: "HTTP: close socket", s,
467: "to retry with Access Authorization");
468: (void)NETCLOSE(s);
469: goto retry;
470: break;
471: }
472: else {
473: /* FALL THROUGH */
474: }
475: default:
476: {
477: char *p1 = HTParse(gate ? gate : arg, "",
478: PARSE_HOST);
479: char * message;
480:
481: if (!(message = (char*)malloc(strlen(text_buffer) +
482: strlen(p1) + 100)))
483: outofmem(__FILE__, "HTTP 4xx status");
484: sprintf(message,
485: "HTTP server at %s replies:\n%s\n\n%s\n",
486: p1, text_buffer,
487: ((server_status == 401)
488: ? "Access Authorization package giving up.\n"
489: : ""));
490: status = HTLoadError(request->output_stream,
491: server_status, message);
492: free(message);
493: free(p1);
494: goto clean_up;
495: }
496: } /* switch */
497: goto clean_up;
498: break;
499: #else
500: /* case 4 without Access Authorization falls through */
501: /* to case 5 (previously "I think I goofed"). -- AL */
502: #endif /* ACCESS_AUTH */
503:
504: case 5: /* I think you goofed */
1.14 luotonen 505: {
506: char *p1 = HTParse(gate ? gate : arg, "", PARSE_HOST);
1.17 timbl 507: char * message = (char*)malloc(
508: strlen(text_buffer)+strlen(p1) + 100);
509: if (!message) outofmem(__FILE__, "HTTP 5xx status");
1.14 luotonen 510: sprintf(message,
1.17 timbl 511: "HTTP server at %s replies:\n%s", p1, text_buffer);
512: status = HTLoadError(request->output_stream, server_status, message);
1.14 luotonen 513: free(message);
514: free(p1);
515: goto clean_up;
516: }
1.17 timbl 517: break;
518:
519: case 2: /* Good: Got MIME object */
520: break;
521:
522: } /* switch on response code */
523:
524: } /* Full HTTP reply */
525:
526: } /* scope of fields */
527:
1.3 timbl 528: /* Set up the stream stack to handle the body of the message
529: */
1.17 timbl 530:
1.13 duns 531: copy:
1.17 timbl 532:
1.18 ! timbl 533: target = HTStreamStack(format_in, request);
1.17 timbl 534:
535: if (!target) {
536: char buffer[1024]; /* @@@@@@@@ */
537: if (binary_buffer) free(binary_buffer);
538: if (text_buffer) free(text_buffer);
539: sprintf(buffer, "Sorry, no known way of converting %s to %s.",
540: HTAtom_name(format_in), HTAtom_name(request->output_format));
541: fprintf(stderr, "HTTP: %s", buffer);
542: status = HTLoadError(request->output_stream, 501, buffer);
543: goto clean_up;
544: }
545:
1.3 timbl 546:
1.11 timbl 547: /* Push the data down the stream
1.3 timbl 548: ** We have to remember the end of the first buffer we just read
1.2 timbl 549: */
1.17 timbl 550: if (format_in == WWW_HTML) {
551: target = HTNetToText(target); /* Pipe through CR stripper */
552: }
553:
554: (*target->isa->put_block)(target,
555: binary_buffer + (start_of_data - text_buffer),
556: length - (start_of_data - text_buffer));
557: HTCopy(s, target);
558:
559: (*target->isa->free)(target);
560: status = HT_LOADED;
1.11 timbl 561:
1.2 timbl 562: /* Clean up
1.1 timbl 563: */
1.17 timbl 564:
565: clean_up:
566: if (binary_buffer) free(binary_buffer);
567: if (text_buffer) free(text_buffer);
568:
569: if (TRACE) fprintf(stderr, "HTTP: close socket %d.\n", s);
570: (void) NETCLOSE(s);
571:
572: return status; /* Good return */
1.3 timbl 573:
1.17 timbl 574: } /* read response */
575: } /* load HTTP */
1.1 timbl 576:
577: /* Protocol descriptor
578: */
579:
1.17 timbl 580: GLOBALDEF PUBLIC HTProtocol HTTP = { "http", HTLoadHTTP, 0, 0 };
Webmaster