HTTP2中文件合并是否仍能减少HTTP头部数量？

Great question—this is one of those "hangover habits" from HTTP/1.1 that a lot of developers struggle to let go of, so let's break it down clearly.

首先，你的直觉没错，但HTTP/2已经把头部开销的问题解决得差不多了 #

You're technically right that merging files reduces the total number of HTTP headers sent. But HTTP/2's HPACK header compression completely changes the math here:

• Most request headers (like Host, User-Agent, Cookie, or even Accept) are identical across multiple requests for JS files on the same domain. HPACK caches these common headers and replaces repeated values with tiny numerical indexes instead of sending the full header text every time.
• For unique headers (like If-None-Match for cache validation), their size is already small, so adding a few more of them barely moves the needle on total load time.

In practice, the "savings" from merging to cut down on headers becomes negligible—we're talking maybe a few hundred bytes at most, which is irrelevant for modern networks.

相反，合并文件会带来更大的性能代价 #

The downsides of merging JS files in HTTP/2 far outweigh any tiny header savings:

• Cache invalidation disaster: If you update just one small piece of code in your merged bundle, every user has to re-download the entire file—even if 90% of it hasn't changed. With separate files, only the updated file needs to be re-fetched, leveraging browser caching far more efficiently.
• Lost parallelism and progressive execution: HTTP/2 uses multiplexing to send multiple files over a single connection simultaneously. Smaller files can start being parsed and executed by the browser as they finish downloading, whereas a single large bundle has to fully download before the browser can do anything with it. This is especially impactful for lazy-loaded code or critical path resources.
• Unnecessary bloat: Merged bundles often include code that's not needed for the current page (e.g., a checkout component on a homepage). This increases your initial load weight, hurting metrics like Largest Contentful Paint (LCP) that directly affect user experience and SEO.

例外情况：极小、紧密耦合的代码片段 #

If you have a handful of tiny, universally used utility functions (think < 1KB total when combined), merging them might make sense—here, the tiny header savings could be slightly more than the overhead of multiple small requests. But this is a rare edge case, not a general rule.

最终结论 #

In HTTP/2, forget about merging JS files for header savings. Focus instead on:

• Splitting code into logical, cacheable chunks
• Using tree-shaking to eliminate unused code
• Optimizing individual files with minification and compression

内容的提问来源于stack exchange，提问作者Richard