Hacker News Reader

Comment by pacifika | Hacker News Reader

I wonder why they wouldn’t take the address at the time of payment so it’s not needed on the actual letter. That seems more convenient. Backwards compatibility?

View on HN

When we started our own parcel service in 2015 we thought that everything that is needed on the actual parcel is the unique ID of the parcel (generated by a system reminiscent of twitter's snowflake and intentionally printed with digits shuffled around as to increase the chance of the prefix being unique). Pretty quickly we found out that various operational concerns need additional data on the label, with routing info (for manual pre-sorting on the depots) and recipients phone number (the courier locates the parcel in the trunk by that) being pretty important.

Also, cool design of the label is one thing, but on both laser and thermal printers the resolution repeatability is much better in the direction perpendicular to the paper travel, so you do not want to do cute design things with vertical barcodes. And you want to have a huge margin below the barcode if it is aligned to the bottom of the label as some printers will occasionally get misaligned and part of the barcode will get on the next label… and the imaging scanners in Zebra terminals will happily read half a milimetre high code-128 barcode instead of the correct one on the bottom of the label.

This isn’t a minor detail. As with stored-value cards vs credit cards, the location of the master record(s) has far-reaching consequences for dependant business processes. Any field activity depending on on-demand access to a centralised API contains the seeds of its own failure.

The availability of centralized API was not the issue. The issue was that the operators were significantly more efficient when performing the process “by eye” than when supported by some kind of computerized system. For example the on-depot sorting process that we settled on was two-phase: first you sort the parcels either manually by routing codes on the labels or through automated sorting line with the current data and then in second phase the operator scans everything that is supposed to go out of the depot and in that process moves around the ~5% of parcels that got sorted incorrectly, which is incredibly more effective than doing that correctly in one pass.

> The availability of centralized API was not the issue. The issue was that the operators were significantly more efficient when performing the process “by eye” than when supported by some kind of computerized system.

If you think these are different considerations, then you are splitting the wrong hairs. The processes landed on to resolve the conflict are not surprising, but also doesn’t sound like they addressed the underlying architectural conceit.

At a large enough scale you can do most of these processes by automation, except the actual courier delivery. And well, at least UPS has actual identified locations in their vans, so just going by location+barcode might be somewhat practical for them.

Beware of the line of operations research thinking that has latterly tanked Boeing’s engineering reputation, something the rest of us designing layered business processes can all learn from. Complexity is always best handled pushed down to the leaves of the tree. The ideal central body performs reporting and audit, not command and control. Although for some, I suppose the Boeing takeaway might be that the bottom line improved and no-one went to prison.

Not sure about NL, but in Switzerland you can buy those codes with an SMS. I guess many people want to do it quickly, and they might have already written the address of the recipient. So just getting a code it much faster, you don't have to enter the address somewhere.

I think the main issue is that regular post has addrrss info of the recipient. So the whole logistics chain is geared towards that. If you remove the need for address info for a (small) subset, you'd have to have two logistics chains somewhat intertwined with each other...

If the delivery model can’t survive an apocalypse, collapse of civilisation, or interruption to the rule of law/network connectivity, then you have no business running a postal service.

But then how would Kevin Costner be able to make a movie out of it???

One may jest but the subtext of The Postman (particularly the David Brin novel, which I can recommend) is analysing which pre-apocalyptic institutional operating models are predisposed (or otherwise) to being rebooted/reconstituted even after the loss of all staff and centralised records, which is a definition of business continuity and process resilience I certainly intended to be in scope of the remark above.

Which is to say, Kevin Costner did make a movie about it. Just not a good one.

In practice though you can simulate/stimulate most of the failure mode conditions of central control dependencies by parking a delivery truck underground, or addressing something to a farm.

If mail gets lost for longer than 5 days, it won’t be known who it goes to, since these codes are reused.

Then you can write any code down and wait 5 days to send it

Probably the sender has to be linked to the current code for it to be accepted

You can, but an invalid code is treated as if there was no postage stamp on it at all. So your letter will be returned or the receiver is asked to pay.

A code that's past the validity date is probably considered an invalid code as well.

Or, you know, the mail man?

Clearly mailmen would just be expected to use their phones to decode the addresses on demand, because everything is supposed to require a phone and an internet connection these days.

A bit more realistically, the address could be printed on a sticker by the sorting machine, the way address redirections are handled.

For a mail/parcel carrier relabeling anything is (somewhat surprisingly) a significant added cost (and there is not that much of margin), so they will do everything in their power to not do that unless absolutely necessary.

PostNL relabels approximately all parcels addressed to me (I almost always have to change either the delivery date or location to one convenient to me), completely free of charge.

Free of charge to you and free of charge to the sender. Which is exactly my point, for PostNL it is a sunk cost and if that will happen for significant amount of their consignments it would be an issue.

> A bit more realistically, the address could be printed on a sticker by the sorting machine, the way address redirections are handled.

Fun fact: PostNL already does this with 100% of the mail processed. The first step in the sorting process is to OCR the address, generate a "KIX" barcode[0], and print that on the letter using special ink if not already present. It's not intended to be user-readable, but anyone with a smartphone can trivially decode the unique destination address from it.

[0]: https://nl.wikipedia.org/wiki/KIX-code

Reply on news.ycombinator.com