⟦7143195c6⟧

TextFile

TODO list for next release.

Merge Charlie's and Kinetics ie.c to get the best version.

Need makefile.68, Kinetics should test on their IS box.

Use technique outlined by Kinetics to reallocate approximately
4 more pbufs from area currently wasted in RAM.

Include some type of date or version stamping mechanism so we can
tell what version various kboxes currently have.  Ddt could
also do some checking to ensure the symbol table corresponds
to the code in the kbox.

Implement configuration flag discussed in last section of 
GATEWAY CONFIGURATION section of ip.at document.

Invent another class of route table entry that looks at the upper
byte only of the appletalk net number.  Then we could have for example
one route table entry in the Stanford table that points to all
nets associated with CMU.  Obviously this requires coordination in the
assignment of the upper byte values:  an appletalk number czar.
When NBP BrRq's are sent to such an 'area route', they 'break open'
at the destination end to hit all appropriate zones.

With the current one packet arouteTuple scheme, only 64 anets can
be accommodated.  Do we need or want to expand to handle more?
Would require multi packet arouteTuples.

Have a config value that is used for testing:  the value would be
the percentage of packets that are simply discarded upon reception(!)
This would allow an easy way to verify error and flow control schemes
in the end to end host implementations.

Add some code to nbpback() that prohibits more than 3 lookups of the
same address within a 30 second period.  This is to prevent 
unkind Mac applications from sending infinite numbers of NBP requests
(each causing many broadcasts) at one second intervals.

Similarly, CAP and KHOST libraries should prevent infinite or
unreasonable use of NBP requests.

Since it makes no sense to use the dummy 3 byte LAP header (it is
just there for byte alignment anyway), use those 3 bytes as a
'sanity check' of valid appletalk-on-IP packets.  In the gateway
it currently stuffs lapdst=0xFA, lapsrc=0xCE, laptype=2 in these
cells.  The CAP/KHOST libraries should also set and check for
these values and discard any appletalk packets which dont
contain them.

Make (minor) mods to KHOST libraries so that they can speak to
a KIP kbox.  This would allow Centram (TOPS) and other companies to relink
their packages and offer a KIP version.

Since we are down to about 18 pbufs now, watch stats to be sure
this is enough.  Get doc from Kinetics on how to add another
RAM chip to the kbox;  (unfortunately adding more RAM precludes
adding a BOOTP PROM, since they both use the same socket(!?))

Modify NBP ARP as per Mark Sherman's suggestion to allow option
of using a plain Mac NBP impleentation as a gateway.  (You can already
use a plain Mac as an IPGATEWAY request server).

Config switch to prohibit sending off-campus IP's.  This could be
turned on in kboxes in dorm situations to prevent internet
'crackers' from causing havoc.

Clarify use of IP broadcasts by MacIP programs.

MacIP address assignment service works fine for medium and small scale
installations, but probably not for large sites such as dorms where you
would want thousands of potential users.  For this case I suggest the
following:  have one or more small UNIX boxes (on ethernet) near the
dorm environment, plus the Kinetics box(es).  Then write a simple
appletalk ATP or ASP driver for the Mac that 'fakes' serial driver
calls.  This (appletalk) serial driver can be used with any of the
popular terminal emulators such as Versaterm or MacTerminal.  Then write
a corresponding CAP or KTALK based daemon for the local UNIX boxes.
Students that wish to connect to a general campus (ethernet only) host
will first log into the local UNIX box, then telnet to the ethernet
only host.  If UNIX boxes are too expensive for this, convince Apple or
Kinetics to make a standalone box (no disk needed) that performs this
multiplexing function.
DataMuseum.dk

DKUUG/EUUG Conference tapes

⟦7143195c6⟧ TextFile

Derivation

TextFile
