[REBOL] Re: [FTP] patch
From: brett:codeconscious at: 18-Nov-2003 18:05
Hi Romano,
> This is my beta version of the patch.
>
> This is what I try to fix:
Great work!
I have looked through your code and I have tried it with my server. It would
be good to use it as a base to build on.
Here's my comments and findings.
---NLST and "Blind" directories
Having looked into NLST more I agree with you - let's just leave it out. If
someone has a desperate need then maybe their situation will show a solution
(custom code perhaps).
Also for "blind" directories better to get what we have working and cleaned
up and deal with the blind problem later if it arises.
--- Parse-files
I believe this function should return none! when parsing fails and a block
when it succeeds. By returning none! an error can be raised within
parse-dir-list. The problem at the moment is if parsing fails, inaccurate
results can be returned without the user knowing. I've confirmed this with
my dodgy test ftp server by emitting deliberately badly formated data (to
simulate an unknown directory listing).
---Data-connect
There's two instances of "error? catch".
---Parse-dir-list
temp: 'list
can be removed by putting the 'list as the parameter directly.
Might be nice to make file-list local to the function.
---PWD and CWD
PWD does not seem to appear at all in network tracing - my patch showed it
coming just after SYST.
I think you may have introduced a bug on CWD. I think that the path should
not be modified if it beings with a "/". I suggest reverting to the original
version of this.
---Unchanged words
These appear to be no different to me: list-check list-file-check
rmdir-check
---Globals
I think parse-dir-list and others can be made local
---"Check you login" bug.
This appears in the original scheme and your scheme, but not my scheme. I
tried to identify what was going on but so far I have not found the problem.
I assume it is to do with connection management and when things are closed.
This is the sequence:
site: ftp://XXXX:[YYY--codeconscious--com]/
print mold read site ; -> SUCCESS
make-dir join site %test-a/test-b/ ; ERROR 550
print mold read site ; SUCCESS
probe read join site %test-a/test-b/ ; ERROR 550
print mold read site ; SUCCESS
probe read join site %test-a/ ; ERROR 550
print mold read site ; ERROR = Failed login to codeconscious...check your
login
probe read join site %test-b/ ; ERROR = Failed login to
codeconscious...check your login
print mold read site ; ERROR = Failed login to codeconscious...check your
login
A strange thing I noticed looking at the trace for the MAKE-DIR was PORT. I
guess I wasn't expecting to see it for make-dir. This was the sequence in
the trace for make-dir:
USER PASS SYST PORT CWD PORT MKD
Maybe some more network logging for the connections will help.
Regards,
Brett.
