Computer tech is self transforming, with new challenges constantly being uncovered and addressed. Problems that baffled computer scientists 25 years ago are mere footnotes, long since conquered. Well, not really.

I'm often taken aback when I realize how many fundamental problems continue to have intrinsically unsatisfactory answers. Some of this is due to growing complexity, some due to art versus science of solutions, some is simple greed. Certain ideas, like the essentially unchanged internal combustion engine, seem to have too much inertia to be replaced by better proposals.

Logins. Authenticating your identity to a computer so you can protect you stuff. :Literally thousands of implementations of this idea exist. Yet still more are continually proposed, because none really work as well as anyone would like.

My current amazement; file transfer. Networks transfer data. While this data isn't necessarily the abstraction we call files, the move file thing idea tends to be implemented soon after the network idea. For the Internet, and home networks, moving files works somewhere above or below TCP. There are many popular protocols and none is perfect.

The venerable FTP has security issues. Windows file standard, an SMB based beast implemented on foreign systems as Samba or CIFS, has a number of issues. (e.g. Samba has a 2GB limit.) NFS has identity issues (all users must have the same internal id, don't get me started.) Anything Apple does seems to work alright for Apple, and no one else. WebDAV...

WebDAV is a pretty clever idea, using HTTP(S) as transfer medium for an FTP like command set with additional meta data goodness. However, like most standards, a complete implementation is hard to come by. Also, it just hasn't gotten enough street cred since its introduction, in spite of being system agnostic and having implementations on Windows, Apple, Linux/Unix(/Apache), and many more.

What do all these systems have in common? They can all be murder to set up. I've set them all up on various systems, mostly Linux and Windows; all have issues, none is clean. Wait, what got me going on this...

FTP! Ancient, simple, FTP. Should be cake on any system, really. Here's my simple, simple, simple config for vsftpd:

anonymous_enable=NO
async_abor_enable=YES
check_shell=NO
chroot_local_user=YES
connect_from_port_20=YES
dirmessage_enable=YES
force_dot_files=YES
listen=YES
local_enable=YES
ls_recurse_enable=YES
write_enable=YES
local_umask=022
# userlist_deny=NO
# userlist_enable=YES
# userlist_file=/etc/vsftpd.user_list -- default location, kill if file here.

Works on most implementations. Allows any establish user to connect. Further fine grained access made by uncommenting the last lines. Of course, most is not all. The Red Hat box fought me for I don't want to know how many hours. The fix was one line.

pam_service_name=vsftpd

Um, yeah, that was obvious. It's all kind of like that. This one is simple. Samba files can be insane. Netatalk is much work for extra crap files in every directory and some buggy behavior. WebDAV, which should be simple, requires an Apache install or a python script from some guy who could be dead.

It's just files! Really, shouldn't there be something that just works with a simple single sign on? Well, there isn't.

From: [identity profile] maccuswell.livejournal.com


What gets me is how wasteful so many of these protocols are. SMB/CIFS is the most chatty network protocol I know, with around 90 RPC connections for a simple directory listing.

WAN links, even high speed WAN links, devastate a lot of file copy protocols for just this reason. Cisco has devices designed simply to cache and augment file copy over remote links, and they ain't cheap.

Perhaps most telling is how Microsoft decided to make and sell Sharepoint instead of fixing their own protocol.
.

Profile

baavgai: (Default)
baavgai

Links

Most Popular Tags

Powered by Dreamwidth Studios

Style Credit

Expand Cut Tags

No cut tags