Because handling those high latency connections will, in turn, increase overhead for both the server itself, and ALL clients connected to the server as it has to broadcast “corrections” to everybody connected to the server. (When players “rubber-band” across the terrain) Which in turn increases latency on the involved client systems as they have to process those corrections, render them, and send (slightly delayed) updated information relevant to the client back to the server.
It can turn into a rather nasty feedback loop where 1 guy with a bad connection manages to drag down 2 or 3 others, who in turn drag down another 6 people, who drag down another dozen, who then pull down another 30 people, and so on. Yeah, Happy Fun Times for all.
The thing about Sharding is it allows for those “duplicated resources” to exist in a different memory space(less of an issue in the era of 64 bit), AND more importantly, it allows that duplication to potentially happen on different hardware. Because sometimes the limiting factor is the physical constraints of the hardware itself, or at least, the efficiency of the code relative to what the hardware can do.
Generally speaking for “short term problems” it’s easier to break things down into smaller chunks and throw more hardware at each chunk to solve a problem than to it is to go about optimizing things in real-time.
This is what Sharding does. It breaks a population apart from being 1 single monolothic chunk into being 2 or more “chunks” instead. Once that population has been divided into different parts, it is then possible to allocate them to different hardware. Divide and conquer, simple as that.
The problem with retail sharding is that they prefer to use the “creamy peanut butter” version of sharding so they can utilize every last nook and crany of existing hardware infrastructure before they resort to using additional hardware. Doing so minimizes their operating costs.
For Classic, there is a case to be made for “creamy peanut butter” in the starting areas. But if they ever try to apply it elsewhere in the game, and that better be exceedingly sparingly, it better be a “super chunky peanut butter” version of sharding that gets used instead.
The retail servers can do their “creamy peanut butter” thing with the idle cycles that is likely to result from this less than optimal configuration on the part of Classic. ![:slight_smile:](https://d38bqls1q93fod.cloudfront.net/en/wow/images/emoji/twitter/slight_smile.png?v=6" ":slight_smile:")
edit: fix broken quote block. Also it should be noted that sharding, by breaking the population into discrete blocks, rather than a monolith, also means any potential impact from “that one guy” who manages to create a the cartoon snowball that takes everybody down with him is also limited to only being able to knock out the one shard, rather than everybody.
![:upside_down_face:](https://d38bqls1q93fod.cloudfront.net/en/wow/images/emoji/twitter/upside_down_face.png?v=6" ":upside_down_face:")