I think the problem is that the  definition of "Sizelim=" is a bit fuzzy.
This keyword was added in 1986 to solve a problem with the redistribution
of ARPA  lists. If  I remember  correctly, the goal  was to  INCREASE the
maximum message  size for lists which  needed it. At the  time, a typical
value for MAILMAXL was 1000, which was too low for some digested lists.
 
To clarify, the purpose of the  (global) MAILMAXL and FILEMAXL limits are
to allow  the maintainer to  tell LISTSERV that  one of the  following is
true:
 
1. There is no point in  attempting to process files larger than XXXXMAXL
   because this  would cause LISTSERV  to run into some  operating system
   quota or  other. That is, it  is known in advance  that this operation
   will  fail, although  it  may  chew quite  a  few  cycles before  this
   happens, and  produce all  sorts of  nasty error  reports. It  is much
   better not  to attempt the  operation. The reason there  are different
   limits for FILE and MAIL is that the latter requires more resources.
 
2. While  LISTSERV's quotas are big  enough to allow the  operation, this
   would be  a disaster  for some other  system component  (typically the
   mail system).  Maybe it  can't handle  files of  that size,  maybe the
   2.4bps  network link  is too  noisy to  transfer a  file of  that size
   before the  usual intermittent  carrier drop,  whatever. Or  maybe the
   other systems on campus just won't accept a message of that size.
 
In practice  people left FILEMAXL alone  and set MAILMAXL to  the largest
message size they  wanted to accept on local mailing  lists and/or to the
largest  message other  systems would  accept. "Sizelim="  allowed larger
values for  lists that really  needed it. Then  I guess what  happened is
that  people discovered  they could  use "Sizelim="  to impose  a maximum
message size  on the list,  and used it that  way. It works,  but doesn't
behave in a very sensible way, since that wasn't its purpose. The logical
thing to do would be to return the  message to the sender. So I think I'm
going to make it  work that way in the next version if  the limit is less
than MAILMAXL.
 
  Eric