[Pan-users] About the flag feature.

[Maurizio Colucci]

Hello,

I am going to argue that there is a subtle conceptual error in the
"flag message" feature that makes it unusable.

Demonstration and suggested fix follows.

--------------------------------------------------------------------

If you have a permanent connection, of course you don't use the flag
feature.

Therefore, let us suppose 

HYPOTHESIS A: you are an offline user = without a permanent
connection.

HYPOTHESIS B: You are looking at the header list ( = message list =
subject list = header pane)

HYPOTHESIS C: you find a subject which definitely raises your
interest.

HYPOTHESIS D: the subject has got some replies.

HYPOTHESIS E: you haven't retrieved any body from that thread yet.


Since by hypothesis the thread definitely interests you, you must read
it.  You have two choices.

1) you flag just the root message

2) flag all the messages.

case 1) you flag just the root message
----------------------------------------

Then you must read it.

- You go online.

- you retrieve flagged bodies.

- you go offline

- you read the root message.

- since by hypothesis the overall subject definitely interests you,
  you would like to read the replies but you can't, because they have
  not been downloaded. You are forced to go online again, spending
  double money. This could have been avoided because YOU ALREADY KNEW
  that the thread definitely interested you.

  FLAWED.


There is another, more serious, problem: ON the next run, you have to
manually flag the new messages (or collapse, unmark and mark the
root). Both ways are terribly impractical! You already know
you want to read all future replies in this thread, so they should be
flagged automatically.


case 2) you flag all the messages
---------------------------------

You have two ways to do that.

2.1) flag the messages one by one.

2.2) collapse the thread and flag the root message.

case 2.1) flag the messages one by one.
----------------------------------------

PROBLEM: too impractical. FLAWED.


case 2.2) collapse the thread and flag the root message.
-------------------------------------------------------

Now you must read the bodies:

 - you go online

 - get new headers (new headers ARE NOT FLAGGED)

 - get flagged messages' bodies (i.e. OLD messages bodies)

 - go offline

 - read the flagged retrieved bodies.

 - you arrive at the new headers in the thread, whose bodies you don't
   have (because they are not flagged). Since by hypothesis they
   interest you, you must connect again to read them. You spend double
   money. This could have been avoided because YOU ALREADY KNEW from
   the subject that you were going to want them.

   FLAWED.


Once again, there is another, more serious, problem: ON the next run,
you have to manually flag the new messages (or collapse, unmark and
mark the root). Both ways are terribly impractical! You already know
you want to read all future replies in this thread, so they should be
flagged automatically.


---------------------------------------------------------


Conclusion
----------

Let us review all use cases in a compact way:

  1) flag just the root message
         FLAWED (expensive and boring to flag each new reply)
  2) flag all the messages
      2.1) flag the messages one by one
            FLAWED (boring)
      2.2) collapse the thread and flag the root.
            FLAWED (expensive and boring to flag each new reply)


Suggested change
----------------

the flag should become an attribute of the THREAD, not of the single
message.  The flag attribute would therefore propagate to new headers.

In that case, you would get new and old bodies with a single
connection:

  - you see an interesting thread.

  - you flag the thread.

  - you go online

  - you get new headers (new replies are automatically flagged)

  - you retrieve the flagged bodies (even new ones)

  - you go offline.

  - you read all messages in the thread, even new ones, which 
     is less expensive.

---------------------------------------------------------

So far we have proved that the change would have an advantage. Now we
must prove that it does not have a drawback.

I can not imagine all possible drawbacks, so I will only discuss
the following:

  You assumed (HYPOTHESIS C) that the subject is unambiguous, so you
  are sure you are going to want to read all the replies. But some
  subjects are ambiguous, and you can't be sure you will want to read
  all the replies. your proposal would have the following drawback:

  - you flag the thread

  - you go online

  - you get new headers (those in the thread are automatically
    flagged)

  - you get flagged bodies

  - you go offline

  - you read the root message of the flagged thread, and discover it
    doesn't interest you. Let us call this thread a FALSE
    INTEREST. You have downloaded so many replies to no
    purpose. Wasted seconds, and money.

ANSWER: you did waste some time, but

1) What is most expensive for offline users is the number of
   connections, not the number of seconds per connection! It is true
   with your method you minimize the lost seconds, but with my method
   you would minimize the number of connections.

   So we should extimate how many seconds you waste with my
   method. Even if the FALSE INTEREST thread has a hundred messages,
   you only loose some seconds. Since FALSE INTEREST threads are RARE,
   you spend few seconds on average.

2) if the root message doesn't interest you, that does not always
   imply the replies won't interest you.