I've been looking at nostr marketplace apps such as Shopstr and Plebeian Market and although they're major steps towards seamless shopping experiences with nostr+Bitcoin, I've noticed that there's a very clear absence of a standardized robust product taxonomy.

We're all familiar with how products are categorized and subcategorized on sites like Amazon and Ebay- there's clear universally accepted hierarchical product categorization e.g. on Amazon you have: Electronics -> Computers & Accessories -> Computer Components -> Graphics Cards -> GeForce RTX 40 Series

This is a fixed taxonomy provided and maintained by Amazon which users are bound to. But on nostr marketplaces clients implement a flat user-generated taxonomy using the "d" tag. For instance:

{
  "kind": 30402,
  "id": "12345.....",
  "pubkey": "12345....",
  "created_at": ,
  "tags": [
    [
      "alt",
      "Product listing: Shirt studs"
    ],
    [
      "title",
      "Shirt studs"
    ],
    [
      "summary",
      "Sterling Silver enamel shirt studs or waistcoat studs"
    ],
    [
      "t",
      "Accessories"
    ],
    [
      "t",
      "Clothing"
    ],
    [
      "t",
      "Collectibles"
    ],

This product is categorized using ["t", "Accessories" ], ["t", "Clothing"] etc. This is problematic for a number of reasons:

1. Tags can't be bundled into a hierarchy that represents parent -> child e.g. Vehicles -> Cars -> Electric Cars
2. Tags can't be given a unique ID (that would prevent splintering due to typos + prevent breaking things in case of renaming)
3. There is no normalized spelling e.g. Art can be either "t": "Art" "t": "art" "t": "ART" or "t": "oil painting" "t": "oil-painting" "t": "oil_painting"
4. No versioning or updates. If a client decides "Computers" becomes "Computers & Accessories", there's no migration path.
5. No metadata possible. You can't attach: description, localized names, associated images/icons, mapping to other taxonomies, SEO keywords etc.
6. Cumbersome discoverability: You can't query “Give me all categories”. You'd need to instead fetch every listing event and then loop through all tag values and build a taxonomy from that.

Is this an issue that's been discussed before?
I'm thinking that using NIP-78 Arbitrary custom app data -k 30078 can be a good solution. It's a kind that already exists and is in use as a kind of "junk drawer".

For example, to create a "Vehicles and Parts" parent category - using Google's free taxonomy - you use the following events - root:

{
  "kind": 30078,
  "pubkey": "<taxonomy-publisher-pubkey>",
  "created_at": 1766271000,
  "tags": [
    ["d", "google:888"],
    ["t", "vehicles"],
    ["alt", "Product category"]
  ],
  "content": {
    "name": "Vehicles & Parts",
    "description": "Products related to vehicles, including complete vehicles and their parts and accessories.",
    "parent": null,

    "i18n": {
      "en": "Vehicles & Parts",
      "he": "רכב וחלקים",
      "es": "Vehículos y repuestos"
    },

    "mappings": {
      "google": "888",
      "wikidata": "Q334166",
      "schema_org": "Product"
    },

    "deprecated": false,
    "source": "google"
  }
}

Child:

{
  "kind": 30078,
  "pubkey": "<taxonomy-publisher-pubkey>",
  "created_at": 1766272000,
  "tags": [
    ["d", "google:916"],
    ["t", "vehicles"],
    ["alt", "Product category"]
  ],
  "content": {
    "name": "Vehicles",
    "description": "Self-propelled or towed conveyances used for transportation.",
    "parent": "google:888",

    "i18n": {
      "en": "Vehicles",
      "he": "כלי רכב",
      "fr": "Véhicules"
    },

    "mappings": {
      "google": "916",
      "wikidata": "Q42889"
    },

    "deprecated": false,
    "source": "google"
  }
}

and so on.

Classified listing can then reference the category via "a" tag:

{
  "kind": 30402,
  "pubkey": "<seller-pubkey>",
  "created_at": 1766273548,
  "tags": [
    ["d", "product_1766262301910_rdbg0"],
    ["title", "Used Electric Hatchback"],
    ["price", "2500", "GBP"],
    ["stock", "1"],

    ["a", "30078:<taxonomy-publisher-pubkey>:google:919"],

    ["t", "vehicle"]
  ],
  "content": "Reliable used electric motor vehicle in good condition."
}

What do you all think of this? Is this the right direction?