[discard] Use bubble plugin instead of pipedream API for the parser

[sync-by-unito]

As a developer, I want to use the Bubble plugin instead of the Pipedream API so that the performance won't be unstable.

┆Issue is synchronized with this Asana task by Unito
┆Created By: Olga

[sync-by-unito]

➤ Olga commented:

Ethan Wu I've set a temporary due date for this task. Please feel free to adjust it based on the milestone date and help add it to the milestone dependencies.

[sync-by-unito]

➤ Ethan Wu commented:

Olga still in discussion with bubble support. as i mentioned today on slack, the original support personale went afk and didn't respond.

i had to reinitiate the conversation and go through it again.

additionally tammy has shared that we may not go with the bubble parser plugin after all because of the performance issues. i believe the conclusion is we will live with pipedream and than move the function to backend in the future.

but i will still communicate with bubble support as this is something we should understand more.

[sync-by-unito]

➤ Olga commented:

Ethan Wu please continue urging Bubble to address the issue. Note that this problem is not linked to the Bubble plugin or Pipedream API; it's related to changes in URL content. Even if we don't use any parser, we still encounter the content change issue. If this can't be resolved, we might need to implement our own method for uploading content instead of relying on Bubble's URL.

[sync-by-unito]

➤ Ethan Wu commented:

current status of discussion:

bubble support has mentioned ignore_imgix again. i have already followed up with them to let them know that this issue exists even with ignore_imgix parameter added.

waiting for feedback

[sync-by-unito]

➤ Ethan Wu commented:

feedback from bubble

Hi Ethan,

Thank you for the follow-up!

After investigating further I noticed a few things. From your reproduction steps, I can clearly see that the first GET request displays the correct data and a second GET request seems to alter it. Here is a Loom to demonstrate this: https://www.loom.com/share/0cc2fbc833304c98b3f8c415ee9b1680?sid=61089705-f320-447d-bb60-f9c345525dbe ( https://www.loom.com/share/0cc2fbc833304c98b3f8c415ee9b1680?sid=61089705-f320-447d-bb60-f9c345525dbe )

I also noticed that when making a second GET request, 3 additional headers are added (Cf-Bgj, Cf-Polished, and Age) which seem to be related to CloudFlare optimization and what causes the metadata to be stripped. More on that here: https://developers.cloudflare.com/images/polish/#:~:text=Polish strips metadata from images and reduces image size ( https://developers.cloudflare.com/images/polish/#:~:text=Polish strips metadata from images and reduces image size )

Another thing to note, when you download the file directly from the file manager, nothing is changed. Loom to demonstrate this:https://www.loom.com/share/e897b56a684a4e4aaefc6af2fe73f2ed?sid=4981c7ef-a919-45ef-8b6a-cdac29e4e5c6 ( https://www.loom.com/share/e897b56a684a4e4aaefc6af2fe73f2ed?sid=4981c7ef-a919-45ef-8b6a-cdac29e4e5c6 ). So it seems to only occur through a request that happens more than once.

I hope this helps clarify the bahavior you are seeing!!

[sync-by-unito]

➤ Ethan Wu commented:

i followed up with a sharing of the bubble plugin and the file that scott used to test before (girl with camera).

i did find though that the plugin worked for another file.

[sync-by-unito]

➤ Olga commented:

regarding the issue Buuble mentioned related to Cloudflare, is there a way for us to obtain the original image even if we need to access it multiple times? Or should you limit access to the Bubble URL to only once to prevent the metadata from being randomly stripped during subsequent accesses?

[sync-by-unito]

➤ Ethan Wu commented:

I will ask

[sync-by-unito]

➤ Ethan Wu commented:

they will probably respond a little late because january 16 is Martin Luther King Day which is a holiday.

i have already asked the question. i'll let you know their responses.

on another note, here (Comment by @Ethan Wu on Use bubble plugin instead of pipedream API for the parser ( https://app.asana.com/0/0/1206201588229874/1206343368232388/f )) i tried another asset with no geolocation and it seems like we were able to get the metadata with the parser.

[sync-by-unito]

➤ Sherry Chung commented:

After previous discussion with Tammy Yang, we might drop this and using backend API. Remove due date.