What old web development habits would seem strange today?

Tables did far more than display data. Whole page layouts could be nested rows and cells, sometimes with tiny transparent images used to hold spacing or stretch a background. It worked until the window size changed, then one cell could shove the rest of the page out of place. CSS layout tools made that kind of structure much less necessary.

Framesets are a good example. A site could keep navigation in one frame and load the page content in another, so the navigation stayed put while the content changed. That came with awkward behavior around links, bookmarks, and browser history, and it is not the same thing as using an iframe for embedded content.

Supporting Internet Explorer sometimes meant maintaining separate stylesheets or adding conditional fixes just for that browser. Rendering differences were large enough that checking a page in several browsers was routine, not a final polish step. Even basic image transparency could lead to format workarounds.

Floats were pressed into service for columns long before modern layout systems were widely available. Then people added clearfix tricks so the container would wrap around floated elements. Floats still have legitimate uses for text wrapping, but treating them as a general-purpose grid system made layouts harder to reason about.

Visual design often started in image-editing software. Developers sliced a mockup into small images and rebuilt the page with tables, background images, and spacer graphics. A lot of the effort was translating a fixed picture into markup that survived different screens, rather than describing a flexible layout directly.