NOTES ON COLD LOAD PROBLEMS:

1. Booted cold load band on processor with Excelan board available.  Why
did we get configured to use it? This caused TIME:INITIALIZE-TIMEBASE
and the first file transfer to fail, with no indication other than "host
not responding". Note that Excelan driver isn't loaded in cold, and
(NET:COLD-LOAD-CONFIGURE) just sets up share and loopback.  (It does
SET-PROCESSOR-OWNING-ETHERNET to :FIND - is that enough to cause the
problem?)

Figuring this out was a little tricky. This pointed up the need for a
NET:STATUS type of function, suitable for calling in the cold load or
random listener,that would print the current network status.
Things to display:

  a. Enabled boards, drivers, protocols
  b. Routing stuff, like CHAOS:LIST-ROUTE-TABLE etc.
  c. Summary statistics
  d. Optionally, DESCRIBE-HOST and (ARP:GET-ADDRESS-INFO on a particular
     host, so we can see why we're losing in each case.


2. Two DEFUN's in SYS:COLD;FIRST are weird: defun of (:property special
special-form-bootstrap) seems to be (are they?) generating a (proper?)
:SOURCE-FILE-NAME property for 'SPECIAL, but the 'UNSPECIAL property is
a munged pathname, "DJ:LISPM;NIL".  Both lead to "...redefining"
messages for the SPECIAL-FORM-BOOTSTRAP while cold-loading.

3. Why is SYS:LAMBDA-DIAG;DIAG-SYSTEM recompiled when OUTER-SYSTEM is
loaded in the cold load world??? ...Is there a mini-LAM subset that we
could load for 2x2 use, without all the voluminous stuff?

4. (SI:QLD) should do PRINT-HERALD before asking for other systems to
load.  And - is that code proceedable, restartable?  Not that I've made
a typo, but if someone does, do they lose big time?

5.  Why aren't early system macro files such as DEFMAC and LMMAC, loaded
in the cold load world?  (Indeed, why isn't the SYSTEM-MACROS system
used for anything?) Yeah, the definitions are already expanded in the
loaded code, but what about potential trouble with initializations,
EVAL's, etc.?  I even got errors in the interpreter, e.g. calling
SECOND.

6. When the network problem occurred, INITIALIZE-TIMEBASE asked me for
the time.  In the cold load, couldn't it call the mini-server and get
the time?  ...Wouldn't that be better anyway?  ...Or with SHARE, do we
get the same effect?