I've been doing BCHS with my computer  group people as a way
of nominally  doing  web but it moreso being a bunch of cool
little  C exercises.   To be fair,  the B, H, and S all  got
substituted,   and I've been looking  slantways  at  the  C.
(netbsd,   netbsd's  bozohttpd,  and I'm not at a  scale  or
relationality  to really want to engage in S).  It's kind of
fun.  I never studied  computers,  but I guess  this is what
baby computer  scientists  fiddle with (mine do anyway).  It
does underscore  some of the high level  features  of common
lisp as compares to C. 

(eq    'foo    (getf   'bar   '(frob   ulous    bar    foo))
> T

But how with...  strtok(3) of a strlcpy(3) of a QUERY_STRING
getenv(3)? 

Something  interesting  to me is that this is almost totally
missing  from gopher.   In gopher  mole cgi, you would  also
access an item type 7 search string  from the environ(7)  of
the cgi, but that doesn't  seem intended  to be like web GET
parameters   (sez me).  Instead  - and I think  this is  the
future of networking - I read in a dream that the gopher has
a 'please give me a 3270 terminal' item specifier. 

I guess this is something  like the historical  boxens   smj
gives  us access  to.  I was sadly  too young  to experience
this, even though I'm old enough now (after  you learn lisp,
you  start  aging in both directions  for both  better   and
worse). 

Tangentially,  smj evidently also has lisp machine boxens. I
should brush up on flavors,  which was the lisp machine lisp
object system.  Changing the common lisp object system to be
like flavors  instead again  was a dry example  of utilising
the metaobject  protocol- that's pretty  much all my context
here. 

Alright, I have been rejuvinated  by non-web topics. Back to
considering web BCHS from the gopher.  I guess the reference
is  those  powerpoint  presentations  from  BSD  conferences
around  2016, which I haven't really been through.   I think
that  what,  mariadb   had been substituted   back  in  with
chameleon after being substituted out, and helper frameworks
such   as kcgi  (I have  no idea  what  kcgi  is  actually).
Alright, kcgi is a very complete BSD ecosystem  light weight
(fast)cgi framework,  and probably is the legacy  of BCHS at
this point. 

I liked the nakedness  of BCHS, the feeling of answering  an
http  request  by just printing  Response:   200 OK CRLF  to
standard out. Trying to reproduce that manually in different
places,   where   it's meant  to be the same,  would   be  a
catastrophe.   On the other hand, the fairly  explicit  kcgi
response   while   gaining   crucial   uniformity    through
abstraction   feels like it loses something,  by  having  an
abstraction that is just as heavy as the thing itself.