summaryrefslogtreecommitdiff
path: root/TODO
diff options
context:
space:
mode:
authorrtomayko <rtomayko>2004-02-14 04:46:05 +0000
committerrtomayko <rtomayko>2004-02-14 04:46:05 +0000
commit12efbb65c44d9798c58c9f0afc53c58729943785 (patch)
treef4f8ea09fe07bd9d99647d279ccd895af628982d /TODO
parent96dfac530b70bd919ded7521aa60aa7449298d18 (diff)
downloadurlgrabber-12efbb65c44d9798c58c9f0afc53c58729943785.tar.gz
Mass commit/merge of changes made for urlgrabber restructuring.. See ChangeLog for details.
Diffstat (limited to 'TODO')
-rw-r--r--TODO37
1 files changed, 37 insertions, 0 deletions
diff --git a/TODO b/TODO
new file mode 100644
index 0000000..900d52a
--- /dev/null
+++ b/TODO
@@ -0,0 +1,37 @@
+Soon:
+
+ * Test proxy support with authenticating proxies (http and ftp).
+
+ * FTP proxy testing in general.
+
+ * Evaluate FTP range implementation. Problem here is that we had to
+ duplicate quite a bit of code from urllib/urllib2 to make this work
+ with the current implementation. We need to accept it and tidy it up a
+ bit or pitch it for something better. I've wracked my head trying to
+ find a better way of doing this. Too much urllib2 source is bad for
+ the head.
+
+ * The FTP REST command does not support specifying an end-byte like HTTP's
+ Range header. Currently, we are calling close() on the file object when
+ the last byte of a range is received. Is there a cleaner way of handling
+ this? Need to verify that the FTP session is being shut down cleanly and
+ the connection is closed properly.
+
+ * Handle non-satisfiable ranges. e.g. due to a server not supporting ranges
+ or a requested range being invalid. Right now the result of either of
+ these two cases is undefined and is probably different for HTTP/FTP.
+ Raising an exception seems to be "the right thing to do" in both cases.
+
+ * Implement MirrorGroups. (Michael to add detail)
+
+ * Possibly clean up conditionals checking for urllib2 availability.
+ Assume urllib2 is available.
+
+Later:
+
+ * Need to test usage across multiple threads. grabber.py should be fine.
+ keepalive seems to be the scary piece here.
+
+ * Consider adding timeout keyword arg that would handle setting the socket
+ timeout natively under python 2.3 or using timeoutsocket.py hack if
+ available and using python < 2.3.