It's true that Wikipedia is not paper, and as such, it's perfectly acceptable for categories for use by readers to be of any size. However, stub categories are not for use by readers, they are for use by editors, who have different requirements. One of those requirements is that they can browse categories of a size that is neither too big to easily hunt down articles nor so small as to necessitate looking through dozens of categories. Categories of between 60 and 800 stubs are an optimum size for this. Anything bigger, and the task becomes too daunting. Anything smaller, and there is serious risk of an editor needing to look in a number of categories while working on a similar subject, and also a danger of a category being repeatedly deleted and re-created as it is emptied and new stubs are made.

A smaller threshold for stub category creation would also lead to a likely proliferation of stub categories, increasing the workload of stub-sorters, who already have to monitor and transfer stubs into several thousand stub categories. Given the number of permcats that have 60 or more articles as a proportion of the total number of permanent categories, the number of stub categories would be likely to blow out into the tens of thousands very rapidly.

Note that the thresholds apply even where it seems likely that a stub category would eventually have the required number of stubs. A possibility or likelihood of there eventually being a certain number of stubs is no guarantee that such stubs will eventually exist – the articles may never be made, or may be made fully-formed at beyond stub size. For this reason stub categories are always created based on the number of currently existing stubs.

Briefly, though it's not a policy, it is a strong recommendation, since it allows the people who work with stubs the most to know what types are likely to be created and to double-check there aren't any problems with them before they are made (it's far easier to make stub types right the first time than try to fix them afterwards!)

Given the enormous number of stub types which exist on Wikipedia, it's also useful for stub types to be arranged as logically as possible so that the types and names of them are clear to those who regularly sort stubs. This includes things like naming stub templates unambiguously and according to a standard naming scheme, making sure that stub category names are as analogous as possible to existing permanent category names, and trying to ensure that stub types are not split in such a way that we end up with a small number of stubs that would require an "everything else" category (see the comments on splitting geography stubs in the next section).

The only way to keep some form of control over this process is to ask for proposals prior to creation. This allows people the opportunity to vet the stub types for any possible problems that may emerge, and to suggest alternatives that might more effectively cover the same ground while being more in keeping with existing stub splits. It also allows for the possibility of contacting any subject-specific WikiProjects to gain input from experts in the particular fields covered by the stubs if there is some doubt as to the best way to split them.

It may seem overly-bureaucratic to do this, but the reasons should be clear – it is the best way to get stub types right from the outset, and it is also the only way to try to keep any control of the available stub types. Without that control, the job of stub-sorting becomes next to impossible.

Sometimes, an editor will suggest splitting a group of stubs into subdivisions in a way that is generally not supported by WikiProject Stub sorting. Though the lack of support for what, initially, seems a sensible split is often surprising to the proposer – especially when that proposer is doing so on behalf of a WikiProject – there are usually very good reasons for not splitting stubs in this particular way. Often this is because of the overall inclusiveness of stub-sorting, i.e., it needs to be able to successfully cover the entirety of Wikipedia, and often this will be at odds with the wishes of one specific WikiProject. The following are some examples of splits which are unlikely to gain much support, even though on the face of it, they appear sensible:

• Splits of sportspeople by team: Sportspeople are, in general, split by sport, nationality and – if further splits are needed – by era. While it may be useful for editors to work on articles on players for a particular team, consider how many stub templates it might be necessary to have for a player with a long career. In sport – particularly in modern sport – a player may play for half a dozen or more teams during a career.
• Splits of biographies by city or subnational region: As with the sportspeople, above, it is quite possible for a person to live in several places during their life and to for their biography to be relevant to all of those places. With a few notable exceptions (such as politicians who represent a specific constituency), biographies are split primarily by nationality and occupation, but not by region.
• Splits of biographies by ethnic group: With rare exceptions (African American, Jewish, Roma), biographies are not split by ethnic group. The reasons for this are three-fold: firstly, it is often difficult to get unequivocal evidence for a person's ethnicity, and even when we can, it is often fairly mixed; secondly, as with independentist regions (below), ethnicity opens up opportunities for edit-warring, especially in cases where there is a strong link between the ethnic mix and politics of a country (e.g., Kurdish Iraqis and Tamil Sri Lankans are simply stubbed as Iraqis and Sri Lankans respectively); thirdly, in many cases, a person's ethnicity is less defining from the point of view of an editor working on a biography than either nationality or occupation. That is, a person will be less likely to achieve fame because of their race than because of where they were and what they did.
• Splits of historical articles by former nation: Stubs are generally, with a few notable exceptions, split by current national boundaries. The reason for this becomes obvious when looking at the history of national boundary changes in Europe in the last 1000 years: we would require separate stub types for literally hundreds of tiny nation states that – to the modern eye – seem to have winked in and out of existence very rapidly. For the sake of ease of navigation, only certain prominent or long-lasting former nations have stub types (among them the Soviet Union, the Roman Empire, the Ottoman Empire, and Ancient Egypt). Even in these cases, stubs relating to articles about definite locations (e.g., geography stubs, building and structure stubs) are always assigned according to current national borders.
• Splits of stubs by independentist region: Generally, these are not done, for the reason that edit wars are likely. Whereas an edit war on an article is usually relatively easy to resolve and only affects that one article, an edit war on a template or category can have far-reaching effects. In general terms, stub types are only split by widely-recognised currently-existing countries and regions. Thus, for example, Cyprus-stub covers the entirety of the island, both the recognised Greek Cypriot state and the less-recognised Turkish Northern Cypriot state.
• Splits of geography stubs by feature: Whereas an individual WikiProject on, say, rivers, can easily divide the articles it finds into rivers and non-rivers, and not have to worry about the non-rivers, a Wikipedia-wide project like WikiProject Stub sorting has to be able to group all articles and not have some which are simply categorised as "everything else". This would be inevitable if geographial stubs were split by feature. The whole planet can be easily split by geopolitical division, and thus this is the easiest way to sort stubs. Where further sorting is necessary, this is done on a subnational basis using local top-level geopolitical divisions (e.g., states, provinces). Apart from being easier for stub sorting to divide by provinces than to split out villages, lakes, mountains, etc., there are also practical benefits for editors. Most editors are more likely to know about their own local environment; it is more likely that an editor in, say Pakistan, will know about other places in Pakistan. As such, dividing by location is the most sensible method of sorting.

In each of these cases, talk page assessment templates are a more sensible way to go; if there is sufficient interest in editing articles on a specific subject not covered by the Wikipedia-wide stub system, then it is likely that a WikiProject or Taskforce will exist to cover that subject. Using a talk-page banner will allow that project to assess and sort all the articles relating to that subject, stubs and non-stubs both. Thus, for instance, WikiProjects or Taskforces on Manchester United, people from Alabama, Celts, the Sparta, Abkhazia, and glaciers could each group their articles using talk-page banner templates to respectively avoid one of the guidelines above. In doing so, they remove any need to rely on specific stub templates for their area of interest.

As mentioned above, and explained at Wikipedia:Stub and Wikipedia:WikiProject Council/Guide#Article tagging, Assessment templates are often a far more useful tool for WikiProjects than stub templates, and don't have the limitations of specific subject requirements or threshold limits. Assessment templates are placed on article talk pages, where they are less likely to be seen as controversial (the placing of stub templates on controversial articles has frequently been a source of edit-warring). They allow all articles within a topic area to be assessed and catalogued by a project, not just stub articles. They allow an indication to be made of exactly what work needs to be done on an article. They also allow workgroups that are subgroups of WikiProjects to have their own specific templates that are better suited to their tasks.

Assessment templates are often used in parallel with stub templates - one for use by a specific WikiProject and monitored by that project, the other for use by Wikipedia editors overall and monitored by Wikipedia:WikiProject Stub sorting. This means that occasionally an article will be regarded as a stub by one project's standards but not by the other, though this should cause no great concern to either. The important thing in these cases – and in every case where either template is used – is that some form of assessment has been made which can be used by both editors interested in a specific field and general editors overall, and each can find the article more readily if it needs expansion.