LSTSRV-L Archives

LISTSERV Site Administrators' Forum

LSTSRV-L

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Topic: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Eric Thomas <ERIC@LEPICS>
Wed, 30 Aug 89 21:59:56 GMT
text/plain (275 lines)
    Description of the changes from release 1.6a to 1.6b of LISTSERV
     --------------------------------------------------------------
                            August 30th, 1989
 
                          *********************
                          * IMPORTANT WARNING *
                          *********************
 
A severe security exposure has been  discovered in LISTSERV, and fixed in
release 1.6b. In  order to protect the integrity of  sites which have not
yet installed release 1.6b, and which may  be unable to do so in a timely
fashion for local reasons, LISTSERV coordination shall disclose NO AMOUNT
of  information   regarding  that  exposure.  LISTSERV   maintainers  are
explicitly  asked not  to discuss  this security  problem on  the network
(and, in  particular, on public lists)  in any way that  might facilitate
the discovery of the  nature of the "hole" by people  who were unaware of
it. Technical discussions which would not provide any "hint" to potential
hackers are of course allowed.
 
****************
* Requirements *
****************
 
LISTSERV  release  1.6b is  supported  on  any hardware/operating  system
configuration supporting release 1.6a.
 
Release 1.6 of  LISTSERV is available only to BITNET  and NetNorth sites,
for  political reasons.  EARN  sites  now receive  support  for the  EARN
version of LISTSERV  from Turgut Kalfaoglu <TURGUT@TREARN>,  to whom EARN
users should address any and all inquiries regarding LISTSERV.
 
*****************
* Compatibility *
*****************
 
Release 1.6b  is fully compatible  with 1.6a,  with the exception  of the
behavioural changes noted under the  headings "Solution to the problem of
CONCEALed users on DISTRIBUTEd lists",  "Handling of moderated lists" and
"Changes in default list header options".
 
Release 1.6b is to be installed directly on top of the base 1.6a version.
 
***************
* Minor fixes *
***************
 
Following a user  request, a brief description of all  the "small yet not
unimportant worries" that have been fixed in each new release will appear
in the  corresponding "Description of  changes" document, along  with the
affected  releases, if  known ("All"  meaning "any  release in  which the
facility is available", ie the bug existed since the very beginning).
 
Releases
affected    Description of the problem
--------    --------------------------
  All       REXX error  if user specifies  database name >  8 characters.
            Fixed.
 
  All       Transferred files  from RSCS  (destination node  unknown) not
            always detected. Fixed.
 
  All       LSVBESTP (called to "allocate" new subscribers to the closest
            peer, among other things) does  not use the same algorithm as
            DIST2,  thus  causing  some  problems  to  sites  with  local
            Internet  gateways  (user  allocated  "far  away"  yet  final
            distribution done  by local server). LSVBESTP  updated to use
            LSVDIST2 logic.
 
  All       Mailing reports that there was  no outbound file when several
            non-local  recipients  were present,  but  all  of them  with
            domain address. Changed message.
 
  All       Incorrect  output   on  "Ack=YES"  lists  when   no  outbound
            recipient. Fixed.
 
*****************************************************
* Evolution: Changes in default list header options *
*****************************************************
 
- The default  value of the  "Ack=" keyword,  which controls the  type of
  acknowledgement sent back as a result  of a mailing operation, has been
  changed  from  "Yes" (mail  acknowledgement  from  each peer)  to  "No"
  (single-line message acknowledgement  from the first peer  only) at the
  request  of what  appeared  to  be a  majority  of  list owners.  Other
  possible values are "MSG" (multi-line message acknowledgement from each
  peer) and "None" (nothing at all).
 
- The  default value  of the  "Errors-To="  keyword, defining  who is  to
  receive notification of processing errors related to the list, has been
  changed  from "Postmaster"  to "Owners"  at the  request of  overloaded
  postmasters who simply  could not handle the load  that deleted student
  accounts  was generating  for them,  and  resent having  to spend  time
  changing the list headers and informing the list owners.
 
************************************************************************
* Evolution: Solution to the problem of CONCEALed users on DISTRIBUTEd *
*            lists (problem introduced in release 1.6a)                *
************************************************************************
 
Concealed users signed up to  DISTRIBUTEd lists had complained that, with
the network  performance enhancement introduced  in the DIST2  command in
release 1.6a, their names would become visible to other recipients of the
same  node, which  is  contrary to  the definition  of  the CONCEAL  list
subscription option. This problem could not be "fixed" quickly because it
was a  design problem  rather than  a bug -  only the  originating server
knows which users are "concealed",  whereas it is the distributing server
which builds the mail headers and control their contents.
 
Rather than restoring the previous behaviour of DIST2 to concealed users,
which, apart from requiring non-trivial changes in the DIST2 protocol for
RFC822  mail,  would  not  have   worked  properly  until  all  potential
"originating"  and "distributing"  servers (ie  most of  the 186  servers
presently in  production) have been  upgraded, it was decided  to deliver
BSMTP-type  headers to  all  concealed recipients  of DISTRIBUTEd  lists.
These headers, being  strictly anonymous, will not  disclose the identity
of the subscriber to anybody, and will not require a separate copy of the
message to be  prepared and sent to these  "concealed" recipients either,
thus not  causing any additional load  on the network as  compared to the
"current" (but unacceptable) 1.6a behaviour.
 
The QUERY command  has been modified to reflect the  actual status of the
"Header="  subscription option,  depending  on the  setting  of both  the
"Conceal=" subscription option and of the "Mail-via=" list header option.
The  SET command  has been  modified to  give a  warning message  when an
attempt to  set the "Header="  option to a non-BSMTP  value is made  by a
concealed user on a DISTRIBUTEd list; the change is accepted, but will be
ignored by  the mail exploder, which  will automatically treat it  as the
corresponding  BSMTP  value  (ie  SHORTHDR  becomes  SHORTBSMTP,  FULLHDR
becomes FULLBSMTP).
 
No action is required from the user, the list owner nor the postmaster to
activate this change.
 
***********************************
* Enhancement: The new TO command *
***********************************
 
A new  "prefix" command, similar  to the  MAIL command (which  causes the
output from  the remainder  of the  command line to  be returned  as mail
rather than interactive  messages - try MAIL INFO for  example), has been
implemented to make it possible to redirect the output of a given command
without having to submit a job file with a "Reply-To=" keyword in the JOB
card. The syntax of this new TO command, which can be freely combined
with the MAIL command, is the following:
 
    TO userid<@node> command-text
or
    TO "user1<@node1> user2<@node2>..." command-text
 
The  default nodeid  is  that  of the  command  originator. For  example,
issuing  "TO  OPERATOR   SHUTDOWN"  from  a  service   machine  with  the
appropriate authority will cause LISTSERV  to stop, sending the resulting
message to the OPERATOR userid rather  than the service machine itself. A
"MAIL TO LISTACNT  SHUTDOWN" command would cause  the shutdown statistics
to  be mailed  to LISTACNT,  which could  be a  server whose  task is  to
collect and process these statistics into daily and weekly reports.
 
*************************************************************************
* Enhancement: Notification of error conditions in list mail processing *
*************************************************************************
 
There are two major types of "error conditions" that can be detected when
mail is being processed by LISTSERV:
 
- Errors  causing the  note to  be rejected,  ie its  distribution to  be
  refused,  because   the  note  is   deemed  to  be  a   delivery  error
  notification; the note will "never" be accepted if re-submitted, unless
  it has been previously edited to remove the patterns that triggered the
  "detector".  These used  to cause  a small  note to  be sent  to either
  postmaster, list owner  or originator, with a copy of  the mail file in
  error sent  under separate cover.  This made it difficult  to correlate
  the  small messages  and actual  mail files,  especially for  non-local
  users. All the information is now provided in a single mail file, which
  contains both the  "diagnostic" message and a copy of  the mail file in
  error.
 
- Errors causing the note not to be  processed for a cause which might be
  transient or correctable (archives disk  full, error reading from disk,
  maximum mail  size exceeded, etc). These  used to send a  small note to
  both  originator  and  postmasters,  with  the  actual  file  being  CP
  TRANSFERred to the "main" postmaster.  This was doubly inconvenient, as
  the  main  postmaster  sometimes  had  trouble  correlating  diagnostic
  message file and  mail file in error, and the  originator had simply no
  way to  know which of  the messages he submitted  was the cause  of the
  error, in  case he had  submitted more than  one in a  relatively short
  period. A single  message with both diagnostic and copy  of mailfile in
  error will now be sent to the originator, while all postmasters get the
  same thing and, possibly (depending on  the exact nature of the error),
  the file  is still CP TRANSFERred  to the "main" postmaster  to make it
  easier to "retry" it.
 
Finally, postings which are rejected because the sender is not authorized
to mail  to the list  are now returned to  the originator along  with the
error message, in case he had not kept a copy in his personal files.
 
******************************************************
* Enhancement/Evolution: Handling of moderated lists *
******************************************************
 
The handling of moderated  ("Send= Editor" or "Send= ...,Semi-Moderated")
lists  has  always been  deemed  unsatisfactory  by  a vast  majority  of
moderators, although none  of them had been able to  propose a better yet
general  solution, which  would benefit  the whole  moderators community,
BITNET and Internet, "filter-ers"  and "digest-ers" alike. Before release
1.6b, mail requiring  moderator approval used to be forwarded  'as is' to
the "main"  editor in the  "preferred" file format  for his node,  with a
small accompanying message (under separate cover) stating why he had been
sent the mail file. This inconvenient for the following reasons:
 
1. The small notes,  which had no value as far  as the moderating process
   is concerned, had to be discarded individually.
 
2. The actual mail file might have been mistaken for private mail, if the
   small note was for some reason delayed somewhere in the network.
 
3. Some  mail packages do  not handle  Netdata input files  properly, and
   this was the format usually chosen by LISTSERV.
 
4. Non-BITNET  moderators  were  getting  the note  enclosed  in  a  mail
   envelope, ie with "double headers", which made it difficult to process
   as commands like FORWARD or REPLY would not behave as expected.
 
The  logic of  moderated lists  has been  changed in  release 1.6b  in an
attempt  to solve  the problems  listed above.  Mail requiring  moderator
approval  is now  sent  via  BSMTP as  a  single  message, with  original
unchanged headers, and with a  descriptive text (the "blurb") inserted on
top of the message body. REPLYing to that message allows the moderator to
ask more information from the sender,  and this works for both BITNET and
Internet editors.
 
In case the editor only acts  as a "filter" for incoming messages, rather
than building  the messages into a  digest, he should simply  FORWARD the
message back to the list, without removing the "blurb", after having made
any change he might have deemed necessary. This is now the "official" way
of  approving a  message  for distribution.  LISTSERV will  automatically
delete the "blurb" from the message  when it is received and accepted for
distribution,  unless  of  course  the  editor  has  already  removed  it
manually. This deletion is done in such a way as to preserve any comments
that the editor might have added before forwarding the message (a la Rice
MAIL). That is, LISTSERV will not delete all the text from the top of the
message  but only  what it  has identified  as being  the "blurb"  it had
inserted itself.
 
Furthermore, if 'Resent-' tags are found in a message sent to a moderated
list and  coming from an  editor, an  'Approved-By:' tag pointing  to the
sender's (editor's) address  is added, whereas the  generated 'From:' tag
will  point to  the person  listed in  the original  'From:' tag,  ie the
person who had  submitted the message in the first  place, rather than to
the editor. Moderators  who make digests have no reason  to forward their
postings  rather than  send them  directly, and  should therefore  not be
affected. Finally, the person who  had originally submitted the note will
get a  copy of the distributed  message regardless of the  setting of his
REPRO  list subscription  option, as  the  editor might  have added  some
comments, removed a  paragraph or otherwise altered the text;  even if no
change has  been made to  the original text, this  note will serve  as an
acknowledgement of the fact that  his contribution has been accepted, and
is  consistent  with  the  behaviour of  previous  versions  of  LISTSERV
wherein, the origin of the distributed  message being that of the editor,
the original  submitter would  automatically be sent  a copy.  The editor
will  also  get his  own  copy,  since the  distribution  acknowledgement
messages generated  by LISTSERV (according  to the setting of  the "Ack="
list header option) will be sent  to the original submitter; the editor's
copy  serves as  an acknowledgement  of  the proper  distribution of  the
message he has approved and forwarded to the list.
 
Moderators who build digests out of the contributions they receive may be
interested in getting rid of the  "blurb" that LISTSERV inserts on top of
the messages, as  they would have to delete it  manually before inserting
the  data into  the digest.  This  can be  done  through the  use of  the
"Editor-header= NO" list  header option, whose use is  not recommended if
you expect to receive any other type of mail in the editor account (ie if
this account is  your personal account rather than a  special one, set up
for  the  sole  purpose  of  receiving  contributions  for  the  list  in
question).

ATOM RSS1 RSS2