Mailing List Archive: 49091 messages
  • Home
  • Script library
  • AltME Archive
  • Mailing list
  • Articles Index
  • Site search
 

[REBOL] Re: problems with url...(and another "escaped" problem)

From: rotenca::telvia::it at: 19-Jun-2002 23:54

Hi Scott, Hingo and all, I have rethinked to the whole question, and now i have more clear mind (i hope).
> The translation is "immediate." This suggests a big at the interpreter > level for url! (and same for file!, but the error is even less likely to > surface due to file naming conventions).
This is Load.
> Other escape characters escape this immediate translation. Witness (with > artificial url): > > >> http://a/ > == http://a/
1) Url! are like file!. 2) The escape sequence is translated like the others. It does not really exist in the string: x: h:/ ; == h:/ type? x ; == url! form x == "h:/ " last x ; == #" " length? x ;== 4 the loaded url does not contain the char sequence " " but a real space. 3) Load convert ALL the %xx values it finds in the string. 4) It is probe/mold which shows the form %xx, because it knows that x is a url! (or a file!). probe x; == h:/ mold x == "h:/ " length? mold x ;== 6 5) But molding a file or url reconverts to the form %xx only the decimals chars: [ 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 34 40 41 59 91 93 123 125 127 160 ] An URL, indeed, requires for some section/schemes that all the non alphanum (non a-z A-Z and little more) chars are escaped. Instead, when you mold an url with a "" char, mold does not show %c8 but "". 6) In the expression: length? to-url "h:/ " ; == 6 the sequence " " is inserted as-is in the string. 7) Escaping the % with %25 and loading the url has the same effect length? h:/%2520 ; == 6
>The answer lies within the http scheme. At one point the "target" >is recreated out of components: >target: next mold to-file join (join "/" either found? port/path >[port/path] [""]) either found? port/target [port/target] [""] > however, the to-file action retranslates it back to > an escape!
Is the [ mold to-file ] expression that re-creates the "%xx" string, but this works only for the already listed chars. Others chars like %c8 does not "re-appear".
> Later, the scheme wishes to check for the need for using a proxy: > > http-packet: reform [http-command either generic-proxy? [port/url] [target] > http-version] > > Here, the generic-proxy version utilizes the "spaced" version; whereas, the > non-proxy version uses the newly created escaped version (that is a path and > target only). A hacked fix would be sure that that the generic-proxy gets a > similarly re-escaped version, which could be done several ways, probably > most cleverly using Ingo's idea.
The problem is that we have no mold support for chars like %c8 and also target is wrong for these chars. A solution could be parsing the string and escape the missed chars or escape all the chars (should be safe). Another workaround is to store always the url in a string and call to-url before open/read/write, but this limits the url! datatypes usage. The final solution should be a url! datatypes (and mold) which knows how to handle the escaped chars in the different section with the different schemes, but i think it is very difficult to write.
> Hope this clarifies the additional mystery.
It helped me. Thanks! --- Ciao Romano