You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm hopping this is not a duplicate issue, since we are multiple to be convinced this is a known bug/limitation, but we struggle to find the corresponding issue.
General problem description
let's say we have a ZIM with a single item at index.html path
in this HTML, we have URLs with fragments, e.g. a user might click on #anchor1 link
if the user has opened the ZIM with the viewer (with the iframe), then
the user is properly redirected the the #anchor1 content
but the URL in the browser stays at something like https://library.acme.com/viewer#my_zim_2024-08/index.html
this is has the adverse effect that users cannot properly bookmark a link directly to #anchor1
this also has the adverse effect that if user reload the page in its navigator, #anchor1 is dropped and he is back to original page
this could have had an impact on page history as well, but it looks like it is working ok
if the user is opening the ZIM without the viewer, everything works correctly:
base URL is something like https://library.acme.com/content/my_zim_2024-08/index.html
when the user clicks on #anchor1 link, URL is updated to https://library.acme.com/content/my_zim_2024-08/index.html#anchor1
this URL can be boorkmarked, the user can reload the page, all will work properly
Why it is becoming important
youtube and freecodecamp scraper (and soon Kolibri, and maybe others) are using Vue.JS ; this means we have a single HTML page and a single JS file, and navigation is handled by JS code.
For proper ZIM operation, we are using the fragment to navigate from page to page (what Vue.JS calls "hash mode" history, see https://router.vuejs.org/guide/essentials/history-mode). This is needed because we really have only one HTML ZIM entry in the ZIM, since we have a single page application.
The problem is however not specific to Vue.JS framework, and is in fact even present without any JS framework, it is just way more visible/important in this scenario.
Should it be too complex to solve this bug, two alternatives to workaround this bug are possible with current usage of Vue.JS, but maybe not with all JS framework and will not work for scenarii without a JS framework.
The main drawback of these alternatives is that it needs modifications of all readers and does not cover all use-cases of the fragment
For this to work, we need a new feature in the ZIM with a "catch-all" ZIM path metadata and a "catch-all" logic in readers. When any path requested is not found by the reader in the ZIM, the reader will return the content of the "catch-all" ZIM path metadata (the index.html in our case).
It obviously has side-effects since we do not have anymore 404 errors for really bad "paths", but in our scenario this is handled by the Vue.JS application which display on its own an error page (just like it does currently if the user pass a bad fragment value).
Alternative 2
In youtube and freecodecamp (and other scrapers), we are already creating additional ZIM entries for every item that we want to be marked as "front" and searchable either in suggestions or full-text search. This is mandatory for proper search operation. These entries have custom indexing data for full-text search, and their content are a very minimal HTML which redirects immediately to the URL with the fragment.
We already know this HTML redirect is kinda a tweak, and should be better supported by the ZIM standard with "real" redirects. Maybe these entries (and their additional future redirect metadata) could be used as well by readers (and especially kiwix-serve) to display proper fragment.
I'm hopping this is not a duplicate issue, since we are multiple to be convinced this is a known bug/limitation, but we struggle to find the corresponding issue.
General problem description
index.html
path#anchor1
link#anchor1
contenthttps://library.acme.com/viewer#my_zim_2024-08/index.html
#anchor1
#anchor1
is dropped and he is back to original pagehttps://library.acme.com/content/my_zim_2024-08/index.html
#anchor1
link, URL is updated tohttps://library.acme.com/content/my_zim_2024-08/index.html#anchor1
Why it is becoming important
youtube and freecodecamp scraper (and soon Kolibri, and maybe others) are using Vue.JS ; this means we have a single HTML page and a single JS file, and navigation is handled by JS code.
For proper ZIM operation, we are using the fragment to navigate from page to page (what Vue.JS calls "hash mode" history, see https://router.vuejs.org/guide/essentials/history-mode). This is needed because we really have only one HTML ZIM entry in the ZIM, since we have a single page application.
The problem is however not specific to Vue.JS framework, and is in fact even present without any JS framework, it is just way more visible/important in this scenario.
Real-world example
See: https://library.kiwix.org/viewer#canadian_prepper_winterprepping_en/index.html ; when you click on any video, the proper video is loaded but nothing changes in browser URL bar. If you want to reload or bookmark a video, you can't.
But https://library.kiwix.org/content/canadian_prepper_winterprepping_en/index.html is working OK.
How other readers are behaving
See kiwix/overview#107 for a complete status.
Alternatives
Should it be too complex to solve this bug, two alternatives to workaround this bug are possible with current usage of Vue.JS, but maybe not with all JS framework and will not work for scenarii without a JS framework.
The main drawback of these alternatives is that it needs modifications of all readers and does not cover all use-cases of the fragment
Alternative 1
In Vue.JS , it is possible to use "HTML5 mode" history, where the fragment is not used but the JS framework intercept the URL to display proper page (see https://router.vuejs.org/guide/essentials/history-mode).
For this to work, we need a new feature in the ZIM with a "catch-all" ZIM path metadata and a "catch-all" logic in readers. When any path requested is not found by the reader in the ZIM, the reader will return the content of the "catch-all" ZIM path metadata (the
index.html
in our case).It obviously has side-effects since we do not have anymore 404 errors for really bad "paths", but in our scenario this is handled by the Vue.JS application which display on its own an error page (just like it does currently if the user pass a bad fragment value).
Alternative 2
In youtube and freecodecamp (and other scrapers), we are already creating additional ZIM entries for every item that we want to be marked as "front" and searchable either in suggestions or full-text search. This is mandatory for proper search operation. These entries have custom indexing data for full-text search, and their content are a very minimal HTML which redirects immediately to the URL with the fragment.
We already know this HTML redirect is kinda a tweak, and should be better supported by the ZIM standard with "real" redirects. Maybe these entries (and their additional future redirect metadata) could be used as well by readers (and especially kiwix-serve) to display proper fragment.
Sample link(entry which could already be displayed / bookmarked today without any ZIM modification (but reader has no idea about how to find them for now): https://library.kiwix.org/viewer#canadian_prepper_winterprepping_en/index/winter_apocalype_survival_must_have_item-z7Ro
The text was updated successfully, but these errors were encountered: