Why everything inside Elemental Chat is being sorted inside the client?

[The-A-Man]

@guillemcordoba

Also, they are not really deleted and cleaned away from the DHT,

Yeah, that so true… In fact, that’s true with even links; they leave ever-lasting waste too…

@guillemcordoba

How are they going to maintain order?

@guillemcordoba

do I need to go over the whole list just to find the one I need to link from?

Sadly, yes… That’s why, especially for use-cases where new entries won’t have to be fit in the middle, it’s a good option. Though not so great for the ones you describe (where new data might have to be fitted within, and not at the ends)…

@guillemcordoba

While interesting and maybe useful in specific cases

Exactly. That’s what I’m also saying; that there are specific use-cases for which Path is a bad option. [And somehow (rather unluckily) I run into them all the time… Haha!]

@guillemcordoba

Another thing is that the UI may very well obscure the details of querying for paths. I think it’s been already 3 projects in which we are doing time indexing via trees of paths, and I think only one of them had explicit time range filter, the other were just infite-scrolling UX patterns like the ones you are describing.

Oh, those are all just a bunch of hacks and fixes; not a cure of the disease itself!

[thedavidmeister]

yeah i think it needs to say ‘links to’

[thedavidmeister]

it’s a tree though with a well formed structure

0 -> all cars above MAX/2
-> all cars below MAX/2 -> all MAX/2 > cars > MAX/4
-> all MAX/4 > cars > 0 -> … MAX/16

etc.

so let’s say the path goes 4 hops down to MAX/16 per bucket ‘o’ price

as a user i want the lowest price, so i can start at MAX/16 > cars > 0

if there are zero results i can ‘zoom out’ to MAX/8 > cars > 0

if there are zero results i can ‘zoom out’ to MAX/4 > cars > 0

if there are results, i can then traverse the graph back down following the low size of each branch that is found

the logic is ‘zoom in and out’ not ‘order and enumerate buckets’

[The-A-Man]

I guess something like this is what you’re talking about:

Sure indeed, at every hop we only have to process a fraction of what we had to in the previous hop; the problem space reduces to half in every step (since, in this example, we have a “binary” one). So, yeah, pretty neat zoom-out, zoom-in.

---

The catch is: we don’t know what the MAX is! And would never know! The market is so dynamic, after all.

Plus neither do we want to visit just about any arbitrary node; we’re only interested in the leftmost branch; so a tree would be an overkill.

[thedavidmeister]

you do know the max because you’re working with unsigned integers that have a maximum value

[The-A-Man]

Yeah, in that sense, you’re right… That’s what I always seem to overlook!