The clause people quote
Public terms for services of this type commonly carry a sentence in which the service describes itself as an intermediary and places the sale itself somewhere else, between the buyer and a seller on a source market. This site does not reproduce the wording of any platform's terms, because it has not read and verified the full text of one. Treat the pattern, not a quotation, as the starting point. A pattern is enough to reason from as long as nobody mistakes it for a verified sentence, and that difference matters when a haul is argued about later.
The pattern is old and it is not unique to shopping agents. Freight forwarders, ticket resellers, marketplaces and payment processors all describe themselves as a layer rather than a party to the underlying transaction. The reason is structural rather than evasive: a layer does not own the goods, cannot warrant their condition, and is not the entity that would answer a claim about them. A service can run an entire pipeline and still not be the seller of anything inside it. That is the entire weight the pattern can carry.
What people do with that sentence is where the trouble starts. It gets quoted as a shield in one direction and as a loophole in the other, and both readings skip the part that matters, which is what the sentence changes about your own records. If the wrapper is a layer, then the facts about a product live upstream of it, and anything you keep downstream carries a dependency you did not choose. Records are editable and dependencies are not, which is why the useful move is upstream rather than into a better folder.
There is also a version of the argument that goes too far. Because the layer is not the seller, some readers conclude that nothing about a stalled haul can be settled anywhere. That does not follow. Stages that were performed by the layer are still stages it performed, and a record of what happened at each one is a record a party can be asked about. Owning goods and operating a pipeline are different things, and both exist at once. Keeping both in view is what separates a fair reading from a convenient one.
The plain reading
The plain reading starts with the address format, which is the one piece of this that can be inspected. In the observed wrapped format, a source field names a market and a second field carries the product number. Across thirty-one wrapped addresses in the 2026-W39 window, every source marker read WD, and WD points at Weidian. That is the entire evidentiary basis for anything this site says about where the goods sit. A window with one marker in it is a small window, and a small window still beats an assertion.
It is a narrow basis and it should be described as one. The format contains a market field, which means other values are possible and probably in use. This site has no observed sample with any other value, so no claim is made about where a Taobao marker or a 1688 marker would lead, and none will be made until such a sample exists. Three market names surround this question. One of them is documented here. The other two are not.
Read literally, a wrapped address is a pointer with a label on it. The label says which market the goods come from, the pointer says which listing on that market, and the wrapper host is a third thing doing a service in the middle. Nothing in that structure transfers ownership of a listing. The listing is the seller's page on the source market, made under that market's rules, and it keeps existing whether or not a wrapper ever pointed at it.
That reading also explains an oddity in this site's own data. All ninety-three records in the last window landed on Weidian, and the count for other markets is zero, which is a statement about the sample rather than about the world. The checks were drawn from publicly visible material, and the publicly visible material carried one marker. A census of what is reachable is not a census of what exists. The same caution applies to any list built from public pages, including the one this article is drawn from.
What it means for a haul
For a haul, the practical consequence is that the fragile part and the valuable part sit in different places. The wrapper is packaging: a layer issued to keep a purchase inside one service, and it expires when that service does. The identifier is closer to property: it belongs to a listing on a market that has its own customers, its own seller, and no stake in whether an agent survives. Packaging dies. Identity tends to survive it. That asymmetry decides where to spend effort, before a purchase rather than after a stoppage.
The second consequence concerns evidence. Photos, measurements and notes produced by the wrapper are products of that layer, and they cannot be carried to another service as though they were facts about the seller's listing. What can be carried is the identity of the listing and anything you wrote down yourself with a date. This is exactly how the sheet column mapper sorts a pasted header row: what is directly reusable, what has to be regenerated elsewhere, and what this site has a written explanation for.
Third, obligations follow the same split. What was bought from the layer was a service: purchasing, receiving, storing, packing, forwarding. What was bought from the seller was an item. When the layer stops, the service stops and the item does not evaporate, which is why warehouse and parcel work could continue while purchasing was suspended in June 2026. The two functions are separable, and reading them as one is what makes a stoppage feel total. Separating them also tells you which questions have an owner and which ones are simply unanswerable.
Fourth, attention belongs upstream. A haul that is being planned is decided mostly by seller-side facts: whether a listing is what it claims to be, whether a size chart is the seller's own, whether a variant is the one pictured. None of those are wrapper facts, and none can be settled by reading a wrapper more carefully. Getting them right before money moves is cheaper than getting them wrong and then looking for somebody to blame. The wrapper was never the right instrument for those questions, and staring at a dead address will not change that.
The action it implies
The first action is to store the source address, not only the wrapped one. Rebuild it once, from the identifier and the market field, keeping the digits exactly as they appear. A rebuilt address is independent of any platform's continued existence, which is the entire point. Rows carrying both addresses can be migrated, checked and compared. Rows carrying only a wrapper cannot be used at all once the wrapper stops. The rebuild is a two-field job, and doing it once is enough for the life of the row.
The second is to record the market marker as its own column instead of burying it inside a pasted address. A column with one observed value in it looks trivial, and it is also the field that decides which source address a rebuild will produce. Writing it down keeps the sample honest as well: a file where every row says WD is describing one market, and its owner should know that before drawing conclusions about three.
The third is to check the platform ledger before choosing a destination, and to read the date on every field rather than the headline. Whether an object takes carried-over addresses, whether it inspects what arrives, which day the entry was last confirmed: those three answers decide the shape of a migration. A name alone decides nothing, and a comparison built on names is a comparison built on nothing. Dates do the work that adjectives cannot, and that is the single idea the platform ledger was built around.
The last action is a habit of language. Say that the listing belongs to a seller on a market, that the service in the middle provided the pipeline, and that the wrapper was packaging around an address that never belonged to it. That sentence is less satisfying than a slogan about who owns what, and it is the one that keeps matching the evidence when the next platform stops answering. Language is not decoration here. It decides which party you write to, and what you ask them for.