Email Parcel Post is Good!

True. However, it's not arbitrary if the ISP says I'm not going to provide
store and forward services for email over 100k (although I usually use 1Meg
as the cutoff). If you want to do something different in your office or
pay for the resources to host some sort of multi-terrabyte mailserver and
the bandwidth to handle all the retries, latent failures, etc., go for it.
Personally, I choose not to accept the headaches for users that don't know
how to use computers.

While "Parcel Post" is arguably valid, there are better methods of achieving
the desired result that are much more net-friendly, user-polite, and efficient.

To further your analogy, parcel post generally accepts packages up to 50#,
nothing more. This is approximately 800 times the weight of a standard letter.
(50 pounds == 800 ounces, and first class letter pricing is based on one ounce

A standard email message is usually less than 32k. 800 times 32k is 25600k,
or about 25 Megabytes. Therefore, using your analogy, it is reasonable to
limit parcel post to 25 Megs. This is giving you substantial bennefit of
the doubt here, as the largest non-attachment email I could find in my inbox
was about 10k (making the parallel limit more like 8 Megs).

Denying other users resources on a store-and-forward system you don't own
by sending large attachments is, in and of itself, arguably a denial of
service attack.


And remember, people don't complain because the local postal system won't
deliver the 150lb anvil that your great-great-great grandpa had dropped on his
head during the civil war. They would also feel bad if they later found out
the the postal worker who had to take the package out of the postal drop
got a hernia from having to take the package out, and because of their
package not only could other people not mail letters, but the letters that
were waiting to be mailed got ruined.

If we are going to compare, does this mean we should start charging extra
for these large messages? If we continue this comparison, a nomral letter
is $0.34 for up to 1 oz. Say a 'normal' attachment (Picture, whatnot) is
1 pound; not heavy, not small enough to just be a letter. It would be $3.45
to send it. That would be 16 times. Now, following your 800x theory, that
would be 50 pounds. To send a 50 pound parcel post would be $36.81. A roughly
1000fold increase.

If we assume, say, a penny per mail (Most of us, of course, bundling the
first, say, 1000 free), to send that 25 MB attachment would cost $10.00.

Would this, then, be considered fair? After all, the post office doesn't
just pick up and deliver our mail, we pay them to do so, on a scale according
to the size, weight, and distance of it. By this logic, we should be able to
do the same and it would be perfectly OK.

(OK, this may be a bit chaotic and disjointed. I've given up caffeine recently
and am a bit skittish for it. :slight_smile:

My point, basically, is that comparing email to the real post office is not
a fair comparison unless we are ready to take on the full burden. Part of that
is to look at each email and consider it in relation to all the others, and
charge the people sending mail according to the resources used.

Personally, I don't think email is an appropriate way to transfer files.
However, I know I am in the minority there, and as such, do not feel the
need to lean on other people to make them stop. I just make sure they know
that sending it to .me. isn't an acceptable method of document transfer.
I have known ISPs who limit, and known ISPs that don't.

Now, what does this have to do with NANOG? Is there a reason to discuss this
here? Is the conversation going to go anywhere other than to more bickering,
and spawn off another off-topic thread in a few days about the merits of
embedded multimedia in HTML pages? No? Then let's drop it. Noone is going
to convert anyone else to their religion, and that is fine. But this thread has
gotten way, way out of hand, and for one that started off-topic, has gotten
to the 'gross' stage. Let us drop this, go have a good weekend, and pick
up Tuesday with a good thread about the merits of IS-IS vs. OSPF or something.