2 mins

I model stuff all the time.

Usually at my desk, often while I’m out walking & nearly every time I’m out on the moto. I think through a problem I’m trying to solve or a thing I want created & mentally interate approaches followed by actual coding. (I’ve actually coded on the bike a couple times using my now-private voice-to-text service… that was wild)

The past few days I’ve rapidly progressed through a series of approaches to the user interface (UI) & user experience (UX) that I want for a filesharing service I developed. I’ll talk more about that at a macro level in another post.

I reached a conclusion earlier today that is a bit disappointing, but still valuable: I won’t be able to produce the UI/UX that I want because of significant performance issues with the core implementation technique.

My previous post - - discussed in fair detail why I wanted to use this approach and concluded with mention of “…finding the boundaries”.

Well, I found some boundaries. They just weren’t what I expected.

When a single image is previewed, the base64 technique is fast & efficient. It is amazingly performant. Once a second and third image is added to the same window, performance degrades rapidly & non-linearly. At image four in a single window, the browser tab locks up. It’s simply exhausted of resources. This was unexpected but is highly reproducible.

I’m stepping back to rethink this at two levels:

  • is this the right approach?
  • is this a problem that needs to be solved?

As I type this, the answer to that second question is unclear.

- jbminn