Re: Call for Community Feedback: Retiring IETF FTP Service

[Keith Moore <[e-mail]>]

On 11/17/20 4:20 PM, Viktor Dukhovni wrote:

> By now I think I also qualify as old timer, and yet I find the claim
> that FTP is simple while HTTP is "complex" far from credible.  If
> anything, the reverse is actually true.  With binary vs. text modes,
> passive vs.  active modes, separate control and data channels, FTP is
> far from "simple".

True, but these days you mostly just use PASV and TYPE I (unless perhaps 
the client detects some baroque OS).   Clever clients try to handle any 
kind of FTP server that might exist, but a client doesn't need to 
support the whole protocol if it knows it's talking to a UNIX/Linux 
system.  (I'm not particularly worried about IETF migrating its servers 
to MCP...)

For comparison, HTTP redirect handling can be somewhat hairy, and add 
TLS on top of that which is quite often necessary these days.   Probably 
most simple clients can get away with using HTTP/1.1 today but who knows 
how long that will last.

> And HTML is not relevant to the discussion.  The
> content type is same (either plain text or HTML) regardless of the
> transfer method.

The relevance is only if you need to parse the HTML, say, to figure out 
the names of files on the server.   Which you often do need to do.   But 
that illustrates a problem with HTTP (sans WebDAV) - the content 
intended for humans isn't clearly distinct from the information needed 
by the client.  (Say you're trying to download new RFCs - the client 
needs to know which <A>'s refer to RFCs and which refer to, say, 
navigation around the site.   It's not rocket science but it is 
annoying, especially given that web sites change from time to time.)

Keith