an Ethernet frame
Think of an Ethernet frame as a single addressed envelope that one machine drops onto the local network. Everything a host sends over a wired LAN — a web request, a file chunk, a video packet — gets wrapped in one of these envelopes before it leaves the network card. The frame is the unit the link layer actually delivers.
The envelope has fixed parts in a fixed order. First a preamble (a repeating pattern of bits, ending in a start-of-frame marker) lets the receiver's clock lock onto the incoming bits. Then comes the 6-byte destination MAC address, the 6-byte source MAC address, and a 2-byte type/length field that says what is inside (for example, the value for an IPv4 packet). Next is the payload — the actual data, between 46 and 1500 bytes (the standard 1500-byte limit is the Ethernet MTU). Finally a 4-byte CRC (the frame check sequence) lets the receiver detect if any bits got corrupted in transit; a damaged frame is silently dropped.
Because the addresses sit right at the front, a switch can read the destination MAC and decide where to send the frame without ever looking at the payload. The CRC only detects errors, it does not fix them — Ethernet leaves recovery to higher layers like TCP. And the 1500-byte payload limit is why a large file is chopped into many frames rather than sent as one.
A frame on the wire: [preamble][dst MAC aa:bb:cc:11:22:33][src MAC de:ad:be:ef:00:01][type 0x0800 = IPv4][payload: an IP packet, 46-1500 bytes][CRC]. The receiver recomputes the CRC; if it differs, the frame is dropped.
Type 0x0800 tells the receiver the payload is an IPv4 packet to hand up to the network layer.
The CRC catches accidental bit errors, not deliberate tampering — it is not a security check. A switch that just forwards by MAC never inspects or verifies the payload.