On Saturday 19 July 2003 8:13 am, Alan DeKok wrote:
> "Alexander M. Pravking" <[EMAIL PROTECTED]> wrote:
> > >   SOME attributes can be set: configuration items.  Some cannot be:
> > > attributes in the request.
> >
> > Alan, could you please describe the difference between them and put it
> > in the FAQ?
>
>   See the 'man' page for the 'users' file.  The document described how
> the 'users' file is processed.  There is NOTHING in there which would
> make anyone believe that attributes can be added to the incoming
> request.

Hmmm...

$ man 5 users
[...]
OPERATORS
       Additional operators other than =  may  be  used  for  the
       attributes  in  either the check item, or reply item list.
[...]
       Attribute := Value
            Always matches as a check item, and replaces  in  the
            configuration  items  any attribute of the same name.
[emphassize]
            If no attribute of that name appears in the  request,
            then this attribute is added.
[/emphasize]
            As a reply item, it has an identical meaning, but for
            the reply items, instead of the request items.

"...if no attribute (exists) then this attribute is added..." pretty much 
contradicts the theory "there is NOTHING in there which would make anyone 
believe attributes can be added to the incoming request", or am I reading 
this wrong?  Or perhaps the text for "+=" is misleading as well when it says 
"always matches and adds the attribute..."

Now, if your going to gripe about the fact I cut off the quote before it said 
"...to the list of configuration items", consider that the "question" posed 
was "SOME attributes can be set: configuration items.  Some cannot be: 
attributes in the request." ... "could you please describe the difference 
between them and put it in the FAQ?"

I think one thing that trips people up is the word "attribute" being used to 
describe different things, and you going ballistic over the fact that end 
users haven't mentally seperated the two meanings out in their head like you 
have.  You may have an almost religious conviction over not helping those who 
do not read, but you've got to realize "reading" alone is half the battle -- 
comprehension must exist for the "reading" to have been of any use.  Further, 
the fact that a particular person doesn't comprehend what you have written is 
not always an indication of that person's abilities, but often indicates what 
is written is not clear.  [like this paragraph :) ]

As much as I hate to say it, as a software & support supplier, you MUST endure 
the "dumb" questions -- in fact, they are often more important than the 
involved questions because they indicate errors in the documentation [and 
writing for people is far harder than writing for computers -- a computer 
will either understand or it won't, and if it doesn't, you get immediate 
feedback (a syntax error); OTOH, a person may understand something OTHER than 
what you meant, but won't give any "feedback" that they might not have 
understood what YOU meant.]

The thing is, you will NEVER get a 100% "correct" document.  The worse part 
about the whole process is that the closer to 100% you get, the more "dumb" 
questions you're going to have to field -- your two standard answers are 
"it's in the document" or "(free)Radius doesn't do that to begin with".  As 
more and more people begin to understand "the document", you get fewer and 
fewer of those questions, leaving you with relatively more "it isn't designed 
for that" questions instead.

[...]
> > Q. What are request attributes?
> > Q. What are config/check items for?
>
>   This is covered in the 'man' page for the 'users' file.

You have an almost dictionary-like definition of these, which may be fine for 
many people, but since this question DOES come up from time to time, not 
everyone sees the connection right off.  Personally, I'd like to see a list 
of ALL attributes along with an indication of "this is an attribute to check 
against" or "this attribute is returned in the response".  [and yes, I know 
the standard answer to THAT question is "buy the book (from amazon so we get 
the kickback)"]  Of course, "all" is a tall order since various NAS makers 
can create their own "vendor-specific" request/response attributes, but 
indicating whether a vendor-specific attribute is a request or response item 
is the responsibility of the vendor.

>   Q: Can radiusd be used to do DHCP?

heh heh heh -- but you know "in a limited sense, it can..." as most people 
think of DHCP as "supplying an IP address" and forgetting that DHCP also 
includes DNS and routing info. :) "for just an IP address", you can specify 
back to the NAS what IP address the client should use, which may satisfy or 
confuse the end user...

>   Q: Can radiusd be used to filter network traffic?

I'm almost tempted to say, "now you're just being silly...", but I see your 
point (and I hope everyone else reading along does too) in that 
packet-filtering is a far cry from user authentication, authorization, and 
accounting, but what about a hybrid solution: using radius to determine what 
users are "granted access" through a firewall [in either direction] (umm, 
that would be the "authorization" part, right? ;) )  In this case, the "NAS", 
per se, isn't a hardware box with modem lines attached, but rather a software 
manifestation running "somewhere" and deciding on how to alter firewall rules 
based on a user id/password combo [and, I presume, source IP address :) ]

[...]
>   The solution is NOT to document each and every dumb thing that
> people try to do. [...] documenting the wrong things is an endless
> treadmill that leads nowhere, and helps no one.

OTOH, pointing out "common misconceptions" has its place as well.  Part of 
what I snipped was your comment that "...people are hyper-afraid of stating 
their end goal..."  I suspect in some (many?) cases a lot of people simply 
don't know what their "goal" is in the first place -- configuring radius is a 
"black art" to them, but their boss got this wild hair growing someplace and 
mandated a "radius server", and the rest, as they say, is history... :)

> ...  I'm NOT willing to add documentation which assumes that the user [is]
> an idiot ...

There is a saying well known in the computer industry [or, at least, it should 
be well known]  "it is very hard to make something idiot proof, because 
idiots are so damn ingenius."  There are those that argue that perhaps we 
"coddle" idiots a bit much [like stamping, "warning -- hot" on a cup of 
McDonald's coffee] and that if we didn't we'd "cull the gene pool" a bit :) 
but let's face it: intelligence in one area doesn't translate to intelligence 
in all areas -- you ARE going to get a "well respected" idiot trying to use 
your program from time to time :)

-- 
Yet another Blog: http://osnut.homelinux.net

Attachment: pgp00000.pgp
Description: signature

Reply via email to